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(dev) \ 304 ((((dev)->hard_header_len + READ_ONCE((dev)->needed_headroom)) \ 305 & ~(HH_DATA_MOD - 1)) + HH_DATA_MOD) 306 #define LL_RESERVED_SPACE_EXTRA(dev,extra) \ 307 ((((dev)->hard_header_len + READ_ONCE((dev)->needed_headroom) + (extra)) \ 308 & ~(HH_DATA_MOD - 1)) + HH_DATA_MOD) 309 310 struct header_ops { 311 int (*create) (struct sk_buff *skb, struct net_device *dev, 312 unsigned short type, const void *daddr, 313 const void *saddr, unsigned int len); 314 int (*parse)(const struct sk_buff *skb, 315 const struct net_device *dev, 316 unsigned char *haddr); 317 int (*cache)(const struct neighbour *neigh, struct hh_cache *hh, __be16 type); 318 void (*cache_update)(struct hh_cache *hh, 319 const struct net_device *dev, 320 const unsigned char *haddr); 321 bool (*validate)(const char *ll_header, unsigned int len); 322 __be16 (*parse_protocol)(const struct sk_buff *skb); 323 }; 324 325 /* These flag bits are private to the generic network queueing 326 * layer; they may not be explicitly referenced by any other 327 * code. 328 */ 329 330 enum netdev_state_t { 331 __LINK_STATE_START, 332 __LINK_STATE_PRESENT, 333 __LINK_STATE_NOCARRIER, 334 __LINK_STATE_LINKWATCH_PENDING, 335 __LINK_STATE_DORMANT, 336 __LINK_STATE_TESTING, 337 }; 338 339 struct gro_list { 340 struct list_head list; 341 int count; 342 }; 343 344 /* 345 * size of gro hash buckets, must be <= the number of bits in 346 * gro_node::bitmask 347 */ 348 #define GRO_HASH_BUCKETS 8 349 350 /** 351 * struct gro_node - structure to support Generic Receive Offload 352 * @bitmask: bitmask to indicate used buckets in @hash 353 * @hash: hashtable of pending aggregated skbs, separated by flows 354 * @rx_list: list of pending ``GRO_NORMAL`` skbs 355 * @rx_count: cached current length of @rx_list 356 * @cached_napi_id: napi_struct::napi_id cached for hotpath, 0 for standalone 357 */ 358 struct gro_node { 359 unsigned long bitmask; 360 struct gro_list hash[GRO_HASH_BUCKETS]; 361 struct list_head rx_list; 362 u32 rx_count; 363 u32 cached_napi_id; 364 }; 365 366 /* 367 * Structure for per-NAPI config 368 */ 369 struct napi_config { 370 u64 gro_flush_timeout; 371 u64 irq_suspend_timeout; 372 u32 defer_hard_irqs; 373 cpumask_t affinity_mask; 374 u8 threaded; 375 unsigned int napi_id; 376 }; 377 378 /* 379 * Structure for NAPI scheduling similar to tasklet but with weighting 380 */ 381 struct napi_struct { 382 /* This field should be first or softnet_data.backlog needs tweaks. */ 383 unsigned long state; 384 /* The poll_list must only be managed by the entity which 385 * changes the state of the NAPI_STATE_SCHED bit. This means 386 * whoever atomically sets that bit can add this napi_struct 387 * to the per-CPU poll_list, and whoever clears that bit 388 * can remove from the list right before clearing the bit. 389 */ 390 struct list_head poll_list; 391 392 int weight; 393 u32 defer_hard_irqs_count; 394 int (*poll)(struct napi_struct *, int); 395 #ifdef CONFIG_NETPOLL 396 /* CPU actively polling if netpoll is configured */ 397 int poll_owner; 398 #endif 399 /* CPU on which NAPI has been scheduled for processing */ 400 int list_owner; 401 struct net_device *dev; 402 struct sk_buff *skb; 403 struct gro_node gro; 404 struct hrtimer timer; 405 /* all fields past this point are write-protected by netdev_lock */ 406 struct task_struct *thread; 407 unsigned long gro_flush_timeout; 408 unsigned long irq_suspend_timeout; 409 u32 defer_hard_irqs; 410 /* control-path-only fields follow */ 411 u32 napi_id; 412 struct list_head dev_list; 413 struct hlist_node napi_hash_node; 414 int irq; 415 struct irq_affinity_notify notify; 416 int napi_rmap_idx; 417 int index; 418 struct napi_config *config; 419 }; 420 421 enum { 422 NAPI_STATE_SCHED, /* Poll is scheduled */ 423 NAPI_STATE_MISSED, /* reschedule a napi */ 424 NAPI_STATE_DISABLE, /* Disable pending */ 425 NAPI_STATE_NPSVC, /* Netpoll - don't dequeue from poll_list */ 426 NAPI_STATE_LISTED, /* NAPI added to system lists */ 427 NAPI_STATE_NO_BUSY_POLL, /* Do not add in napi_hash, no busy polling */ 428 NAPI_STATE_IN_BUSY_POLL, /* Do not rearm NAPI interrupt */ 429 NAPI_STATE_PREFER_BUSY_POLL, /* prefer busy-polling over softirq processing*/ 430 NAPI_STATE_THREADED, /* The poll is performed inside its own thread*/ 431 NAPI_STATE_SCHED_THREADED, /* Napi is currently scheduled in threaded mode */ 432 NAPI_STATE_HAS_NOTIFIER, /* Napi has an IRQ notifier */ 433 NAPI_STATE_THREADED_BUSY_POLL, /* The threaded NAPI poller will busy poll */ 434 }; 435 436 enum { 437 NAPIF_STATE_SCHED = BIT(NAPI_STATE_SCHED), 438 NAPIF_STATE_MISSED = BIT(NAPI_STATE_MISSED), 439 NAPIF_STATE_DISABLE = BIT(NAPI_STATE_DISABLE), 440 NAPIF_STATE_NPSVC = BIT(NAPI_STATE_NPSVC), 441 NAPIF_STATE_LISTED = BIT(NAPI_STATE_LISTED), 442 NAPIF_STATE_NO_BUSY_POLL = BIT(NAPI_STATE_NO_BUSY_POLL), 443 NAPIF_STATE_IN_BUSY_POLL = BIT(NAPI_STATE_IN_BUSY_POLL), 444 NAPIF_STATE_PREFER_BUSY_POLL = BIT(NAPI_STATE_PREFER_BUSY_POLL), 445 NAPIF_STATE_THREADED = BIT(NAPI_STATE_THREADED), 446 NAPIF_STATE_SCHED_THREADED = BIT(NAPI_STATE_SCHED_THREADED), 447 NAPIF_STATE_HAS_NOTIFIER = BIT(NAPI_STATE_HAS_NOTIFIER), 448 NAPIF_STATE_THREADED_BUSY_POLL = BIT(NAPI_STATE_THREADED_BUSY_POLL), 449 }; 450 451 enum gro_result { 452 GRO_MERGED, 453 GRO_MERGED_FREE, 454 GRO_HELD, 455 GRO_NORMAL, 456 GRO_CONSUMED, 457 }; 458 typedef enum gro_result gro_result_t; 459 460 /* 461 * enum rx_handler_result - Possible return values for rx_handlers. 462 * @RX_HANDLER_CONSUMED: skb was consumed by rx_handler, do not process it 463 * further. 464 * @RX_HANDLER_ANOTHER: Do another round in receive path. This is indicated in 465 * case skb->dev was changed by rx_handler. 466 * @RX_HANDLER_EXACT: Force exact delivery, no wildcard. 467 * @RX_HANDLER_PASS: Do nothing, pass the skb as if no rx_handler was called. 468 * 469 * rx_handlers are functions called from inside __netif_receive_skb(), to do 470 * special processing of the skb, prior to delivery to protocol handlers. 471 * 472 * Currently, a net_device can only have a single rx_handler registered. Trying 473 * to register a second rx_handler will return -EBUSY. 474 * 475 * To register a rx_handler on a net_device, use netdev_rx_handler_register(). 476 * To unregister a rx_handler on a net_device, use 477 * netdev_rx_handler_unregister(). 478 * 479 * Upon return, rx_handler is expected to tell __netif_receive_skb() what to 480 * do with the skb. 481 * 482 * If the rx_handler consumed the skb in some way, it should return 483 * RX_HANDLER_CONSUMED. This is appropriate when the rx_handler arranged for 484 * the skb to be delivered in some other way. 485 * 486 * If the rx_handler changed skb->dev, to divert the skb to another 487 * net_device, it should return RX_HANDLER_ANOTHER. The rx_handler for the 488 * new device will be called if it exists. 489 * 490 * If the rx_handler decides the skb should be ignored, it should return 491 * RX_HANDLER_EXACT. The skb will only be delivered to protocol handlers that 492 * are registered on exact device (ptype->dev == skb->dev). 493 * 494 * If the rx_handler didn't change skb->dev, but wants the skb to be normally 495 * delivered, it should return RX_HANDLER_PASS. 496 * 497 * A device without a registered rx_handler will behave as if rx_handler 498 * returned RX_HANDLER_PASS. 499 */ 500 501 enum rx_handler_result { 502 RX_HANDLER_CONSUMED, 503 RX_HANDLER_ANOTHER, 504 RX_HANDLER_EXACT, 505 RX_HANDLER_PASS, 506 }; 507 typedef enum rx_handler_result rx_handler_result_t; 508 typedef rx_handler_result_t rx_handler_func_t(struct sk_buff **pskb); 509 510 void __napi_schedule(struct napi_struct *n); 511 void __napi_schedule_irqoff(struct napi_struct *n); 512 513 static inline bool napi_disable_pending(struct napi_struct *n) 514 { 515 return test_bit(NAPI_STATE_DISABLE, &n->state); 516 } 517 518 static inline bool napi_prefer_busy_poll(struct napi_struct *n) 519 { 520 return test_bit(NAPI_STATE_PREFER_BUSY_POLL, &n->state); 521 } 522 523 /** 524 * napi_is_scheduled - test if NAPI is scheduled 525 * @n: NAPI context 526 * 527 * This check is "best-effort". With no locking implemented, 528 * a NAPI can be scheduled or terminate right after this check 529 * and produce not precise results. 530 * 531 * NAPI_STATE_SCHED is an internal state, napi_is_scheduled 532 * should not be used normally and napi_schedule should be 533 * used instead. 534 * 535 * Use only if the driver really needs to check if a NAPI 536 * is scheduled for example in the context of delayed timer 537 * that can be skipped if a NAPI is already scheduled. 538 * 539 * Return: True if NAPI is scheduled, False otherwise. 540 */ 541 static inline bool napi_is_scheduled(struct napi_struct *n) 542 { 543 return test_bit(NAPI_STATE_SCHED, &n->state); 544 } 545 546 bool napi_schedule_prep(struct napi_struct *n); 547 548 /** 549 * napi_schedule - schedule NAPI poll 550 * @n: NAPI context 551 * 552 * Schedule NAPI poll routine to be called if it is not already 553 * running. 554 * Return: true if we schedule a NAPI or false if not. 555 * Refer to napi_schedule_prep() for additional reason on why 556 * a NAPI might not be scheduled. 557 */ 558 static inline bool napi_schedule(struct napi_struct *n) 559 { 560 if (napi_schedule_prep(n)) { 561 __napi_schedule(n); 562 return true; 563 } 564 565 return false; 566 } 567 568 /** 569 * napi_schedule_irqoff - schedule NAPI poll 570 * @n: NAPI context 571 * 572 * Variant of napi_schedule(), assuming hard irqs are masked. 573 */ 574 static inline void napi_schedule_irqoff(struct napi_struct *n) 575 { 576 if (napi_schedule_prep(n)) 577 __napi_schedule_irqoff(n); 578 } 579 580 /** 581 * napi_complete_done - NAPI processing complete 582 * @n: NAPI context 583 * @work_done: number of packets processed 584 * 585 * Mark NAPI processing as complete. Should only be called if poll budget 586 * has not been completely consumed. 587 * Prefer over napi_complete(). 588 * Return: false if device should avoid rearming interrupts. 589 */ 590 bool napi_complete_done(struct napi_struct *n, int work_done); 591 592 static inline bool napi_complete(struct napi_struct *n) 593 { 594 return napi_complete_done(n, 0); 595 } 596 597 void netif_threaded_enable(struct net_device *dev); 598 int dev_set_threaded(struct net_device *dev, 599 enum netdev_napi_threaded threaded); 600 601 void napi_disable(struct napi_struct *n); 602 void napi_disable_locked(struct napi_struct *n); 603 604 void napi_enable(struct napi_struct *n); 605 void napi_enable_locked(struct napi_struct *n); 606 607 /** 608 * napi_synchronize - wait until NAPI is not running 609 * @n: NAPI context 610 * 611 * Wait until NAPI is done being scheduled on this context. 612 * Waits till any outstanding processing completes but 613 * does not disable future activations. 614 */ 615 static inline void napi_synchronize(const struct napi_struct *n) 616 { 617 if (IS_ENABLED(CONFIG_SMP)) 618 while (test_bit(NAPI_STATE_SCHED, &n->state)) 619 msleep(1); 620 else 621 barrier(); 622 } 623 624 /** 625 * napi_if_scheduled_mark_missed - if napi is running, set the 626 * NAPIF_STATE_MISSED 627 * @n: NAPI context 628 * 629 * If napi is running, set the NAPIF_STATE_MISSED, and return true if 630 * NAPI is scheduled. 631 **/ 632 static inline bool napi_if_scheduled_mark_missed(struct napi_struct *n) 633 { 634 unsigned long val, new; 635 636 val = READ_ONCE(n->state); 637 do { 638 if (val & NAPIF_STATE_DISABLE) 639 return true; 640 641 if (!(val & NAPIF_STATE_SCHED)) 642 return false; 643 644 new = val | NAPIF_STATE_MISSED; 645 } while (!try_cmpxchg(&n->state, &val, new)); 646 647 return true; 648 } 649 650 enum netdev_queue_state_t { 651 __QUEUE_STATE_DRV_XOFF, 652 __QUEUE_STATE_STACK_XOFF, 653 __QUEUE_STATE_FROZEN, 654 }; 655 656 #define QUEUE_STATE_DRV_XOFF (1 << __QUEUE_STATE_DRV_XOFF) 657 #define QUEUE_STATE_STACK_XOFF (1 << __QUEUE_STATE_STACK_XOFF) 658 #define QUEUE_STATE_FROZEN (1 << __QUEUE_STATE_FROZEN) 659 660 #define QUEUE_STATE_ANY_XOFF (QUEUE_STATE_DRV_XOFF | QUEUE_STATE_STACK_XOFF) 661 #define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \ 662 QUEUE_STATE_FROZEN) 663 #define QUEUE_STATE_DRV_XOFF_OR_FROZEN (QUEUE_STATE_DRV_XOFF | \ 664 QUEUE_STATE_FROZEN) 665 666 /* 667 * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue. The 668 * netif_tx_* functions below are used to manipulate this flag. The 669 * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit 670 * queue independently. The netif_xmit_*stopped functions below are called 671 * to check if the queue has been stopped by the driver or stack (either 672 * of the XOFF bits are set in the state). Drivers should not need to call 673 * netif_xmit*stopped functions, they should only be using netif_tx_*. 674 */ 675 676 struct netdev_queue { 677 /* 678 * read-mostly part 679 */ 680 struct net_device *dev; 681 netdevice_tracker dev_tracker; 682 683 struct Qdisc __rcu *qdisc; 684 struct Qdisc __rcu *qdisc_sleeping; 685 #ifdef CONFIG_SYSFS 686 struct kobject kobj; 687 const struct attribute_group **groups; 688 #endif 689 unsigned long tx_maxrate; 690 /* 691 * Number of TX timeouts for this queue 692 * (/sys/class/net/DEV/Q/trans_timeout) 693 */ 694 atomic_long_t trans_timeout; 695 696 /* Subordinate device that the queue has been assigned to */ 697 struct net_device *sb_dev; 698 #ifdef CONFIG_XDP_SOCKETS 699 /* "ops protected", see comment about net_device::lock */ 700 struct xsk_buff_pool *pool; 701 #endif 702 703 /* 704 * write-mostly part 705 */ 706 #ifdef CONFIG_BQL 707 struct dql dql; 708 #endif 709 spinlock_t _xmit_lock ____cacheline_aligned_in_smp; 710 int xmit_lock_owner; 711 /* 712 * Time (in jiffies) of last Tx 713 */ 714 unsigned long trans_start; 715 716 unsigned long state; 717 718 /* 719 * slow- / control-path part 720 */ 721 /* NAPI instance for the queue 722 * "ops protected", see comment about net_device::lock 723 */ 724 struct napi_struct *napi; 725 726 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA) 727 int numa_node; 728 #endif 729 } ____cacheline_aligned_in_smp; 730 731 extern int sysctl_fb_tunnels_only_for_init_net; 732 extern int sysctl_devconf_inherit_init_net; 733 734 /* 735 * sysctl_fb_tunnels_only_for_init_net == 0 : For all netns 736 * == 1 : For initns only 737 * == 2 : For none. 738 */ 739 static inline bool net_has_fallback_tunnels(const struct net *net) 740 { 741 #if IS_ENABLED(CONFIG_SYSCTL) 742 int fb_tunnels_only_for_init_net = READ_ONCE(sysctl_fb_tunnels_only_for_init_net); 743 744 return !fb_tunnels_only_for_init_net || 745 (net_eq(net, &init_net) && fb_tunnels_only_for_init_net == 1); 746 #else 747 return true; 748 #endif 749 } 750 751 static inline int net_inherit_devconf(void) 752 { 753 #if IS_ENABLED(CONFIG_SYSCTL) 754 return READ_ONCE(sysctl_devconf_inherit_init_net); 755 #else 756 return 0; 757 #endif 758 } 759 760 static inline int netdev_queue_numa_node_read(const struct netdev_queue *q) 761 { 762 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA) 763 return q->numa_node; 764 #else 765 return NUMA_NO_NODE; 766 #endif 767 } 768 769 static inline void netdev_queue_numa_node_write(struct netdev_queue *q, int node) 770 { 771 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA) 772 q->numa_node = node; 773 #endif 774 } 775 776 #ifdef CONFIG_RFS_ACCEL 777 bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index, u32 flow_id, 778 u16 filter_id); 779 #endif 780 781 /* XPS map type and offset of the xps map within net_device->xps_maps[]. */ 782 enum xps_map_type { 783 XPS_CPUS = 0, 784 XPS_RXQS, 785 XPS_MAPS_MAX, 786 }; 787 788 #ifdef CONFIG_XPS 789 /* 790 * This structure holds an XPS map which can be of variable length. The 791 * map is an array of queues. 792 */ 793 struct xps_map { 794 unsigned int len; 795 unsigned int alloc_len; 796 struct rcu_head rcu; 797 u16 queues[]; 798 }; 799 #define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16))) 800 #define XPS_MIN_MAP_ALLOC ((L1_CACHE_ALIGN(offsetof(struct xps_map, queues[1])) \ 801 - sizeof(struct xps_map)) / sizeof(u16)) 802 803 /* 804 * This structure holds all XPS maps for device. Maps are indexed by CPU. 805 * 806 * We keep track of the number of cpus/rxqs used when the struct is allocated, 807 * in nr_ids. This will help not accessing out-of-bound memory. 808 * 809 * We keep track of the number of traffic classes used when the struct is 810 * allocated, in num_tc. This will be used to navigate the maps, to ensure we're 811 * not crossing its upper bound, as the original dev->num_tc can be updated in 812 * the meantime. 813 */ 814 struct xps_dev_maps { 815 struct rcu_head rcu; 816 unsigned int nr_ids; 817 s16 num_tc; 818 struct xps_map __rcu *attr_map[]; /* Either CPUs map or RXQs map */ 819 }; 820 821 #define XPS_CPU_DEV_MAPS_SIZE(_tcs) (sizeof(struct xps_dev_maps) + \ 822 (nr_cpu_ids * (_tcs) * sizeof(struct xps_map *))) 823 824 #define XPS_RXQ_DEV_MAPS_SIZE(_tcs, _rxqs) (sizeof(struct xps_dev_maps) +\ 825 (_rxqs * (_tcs) * sizeof(struct xps_map *))) 826 827 #endif /* CONFIG_XPS */ 828 829 #define TC_MAX_QUEUE 16 830 #define TC_BITMASK 15 831 /* HW offloaded queuing disciplines txq count and offset maps */ 832 struct netdev_tc_txq { 833 u16 count; 834 u16 offset; 835 }; 836 837 #if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE) 838 /* 839 * This structure is to hold information about the device 840 * configured to run FCoE protocol stack. 841 */ 842 struct netdev_fcoe_hbainfo { 843 char manufacturer[64]; 844 char serial_number[64]; 845 char hardware_version[64]; 846 char driver_version[64]; 847 char optionrom_version[64]; 848 char firmware_version[64]; 849 char model[256]; 850 char model_description[256]; 851 }; 852 #endif 853 854 #define MAX_PHYS_ITEM_ID_LEN 32 855 856 /* This structure holds a unique identifier to identify some 857 * physical item (port for example) used by a netdevice. 858 */ 859 struct netdev_phys_item_id { 860 unsigned char id[MAX_PHYS_ITEM_ID_LEN]; 861 unsigned char id_len; 862 }; 863 864 static inline bool netdev_phys_item_id_same(struct netdev_phys_item_id *a, 865 struct netdev_phys_item_id *b) 866 { 867 return a->id_len == b->id_len && 868 memcmp(a->id, b->id, a->id_len) == 0; 869 } 870 871 typedef u16 (*select_queue_fallback_t)(struct net_device *dev, 872 struct sk_buff *skb, 873 struct net_device *sb_dev); 874 875 enum net_device_path_type { 876 DEV_PATH_ETHERNET = 0, 877 DEV_PATH_VLAN, 878 DEV_PATH_BRIDGE, 879 DEV_PATH_PPPOE, 880 DEV_PATH_DSA, 881 DEV_PATH_MTK_WDMA, 882 DEV_PATH_TUN, 883 }; 884 885 struct net_device_path { 886 enum net_device_path_type type; 887 const struct net_device *dev; 888 union { 889 struct { 890 u16 id; 891 __be16 proto; 892 u8 h_dest[ETH_ALEN]; 893 } encap; 894 struct { 895 union { 896 struct in_addr src_v4; 897 struct in6_addr src_v6; 898 }; 899 union { 900 struct in_addr dst_v4; 901 struct in6_addr dst_v6; 902 }; 903 904 u8 l3_proto; 905 } tun; 906 struct { 907 enum { 908 DEV_PATH_BR_VLAN_KEEP, 909 DEV_PATH_BR_VLAN_TAG, 910 DEV_PATH_BR_VLAN_UNTAG, 911 DEV_PATH_BR_VLAN_UNTAG_HW, 912 } vlan_mode; 913 u16 vlan_id; 914 __be16 vlan_proto; 915 } bridge; 916 struct { 917 int port; 918 u16 proto; 919 } dsa; 920 struct { 921 u8 wdma_idx; 922 u8 queue; 923 u16 wcid; 924 u8 bss; 925 u8 amsdu; 926 } mtk_wdma; 927 }; 928 }; 929 930 #define NET_DEVICE_PATH_STACK_MAX 5 931 #define NET_DEVICE_PATH_VLAN_MAX 2 932 933 struct net_device_path_stack { 934 int num_paths; 935 struct net_device_path path[NET_DEVICE_PATH_STACK_MAX]; 936 }; 937 938 struct net_device_path_ctx { 939 const struct net_device *dev; 940 u8 daddr[ETH_ALEN]; 941 942 int num_vlans; 943 struct { 944 u16 id; 945 __be16 proto; 946 } vlan[NET_DEVICE_PATH_VLAN_MAX]; 947 }; 948 949 enum tc_setup_type { 950 TC_QUERY_CAPS, 951 TC_SETUP_QDISC_MQPRIO, 952 TC_SETUP_CLSU32, 953 TC_SETUP_CLSFLOWER, 954 TC_SETUP_CLSMATCHALL, 955 TC_SETUP_CLSBPF, 956 TC_SETUP_BLOCK, 957 TC_SETUP_QDISC_CBS, 958 TC_SETUP_QDISC_RED, 959 TC_SETUP_QDISC_PRIO, 960 TC_SETUP_QDISC_MQ, 961 TC_SETUP_QDISC_ETF, 962 TC_SETUP_ROOT_QDISC, 963 TC_SETUP_QDISC_GRED, 964 TC_SETUP_QDISC_TAPRIO, 965 TC_SETUP_FT, 966 TC_SETUP_QDISC_ETS, 967 TC_SETUP_QDISC_TBF, 968 TC_SETUP_QDISC_FIFO, 969 TC_SETUP_QDISC_HTB, 970 TC_SETUP_ACT, 971 }; 972 973 /* These structures hold the attributes of bpf state that are being passed 974 * to the netdevice through the bpf op. 975 */ 976 enum bpf_netdev_command { 977 /* Set or clear a bpf program used in the earliest stages of packet 978 * rx. The prog will have been loaded as BPF_PROG_TYPE_XDP. The callee 979 * is responsible for calling bpf_prog_put on any old progs that are 980 * stored. In case of error, the callee need not release the new prog 981 * reference, but on success it takes ownership and must bpf_prog_put 982 * when it is no longer used. 983 */ 984 XDP_SETUP_PROG, 985 XDP_SETUP_PROG_HW, 986 /* BPF program for offload callbacks, invoked at program load time. */ 987 BPF_OFFLOAD_MAP_ALLOC, 988 BPF_OFFLOAD_MAP_FREE, 989 XDP_SETUP_XSK_POOL, 990 }; 991 992 struct bpf_prog_offload_ops; 993 struct netlink_ext_ack; 994 struct xdp_umem; 995 struct xdp_dev_bulk_queue; 996 struct bpf_xdp_link; 997 998 enum bpf_xdp_mode { 999 XDP_MODE_SKB = 0, 1000 XDP_MODE_DRV = 1, 1001 XDP_MODE_HW = 2, 1002 __MAX_XDP_MODE 1003 }; 1004 1005 struct bpf_xdp_entity { 1006 struct bpf_prog *prog; 1007 struct bpf_xdp_link *link; 1008 }; 1009 1010 struct netdev_bpf { 1011 enum bpf_netdev_command command; 1012 union { 1013 /* XDP_SETUP_PROG */ 1014 struct { 1015 u32 flags; 1016 struct bpf_prog *prog; 1017 struct netlink_ext_ack *extack; 1018 }; 1019 /* BPF_OFFLOAD_MAP_ALLOC, BPF_OFFLOAD_MAP_FREE */ 1020 struct { 1021 struct bpf_offloaded_map *offmap; 1022 }; 1023 /* XDP_SETUP_XSK_POOL */ 1024 struct { 1025 struct xsk_buff_pool *pool; 1026 u16 queue_id; 1027 } xsk; 1028 }; 1029 }; 1030 1031 /* Flags for ndo_xsk_wakeup. */ 1032 #define XDP_WAKEUP_RX (1 << 0) 1033 #define XDP_WAKEUP_TX (1 << 1) 1034 1035 #ifdef CONFIG_XFRM_OFFLOAD 1036 struct xfrmdev_ops { 1037 int (*xdo_dev_state_add)(struct net_device *dev, 1038 struct xfrm_state *x, 1039 struct netlink_ext_ack *extack); 1040 void (*xdo_dev_state_delete)(struct net_device *dev, 1041 struct xfrm_state *x); 1042 void (*xdo_dev_state_free)(struct net_device *dev, 1043 struct xfrm_state *x); 1044 bool (*xdo_dev_offload_ok) (struct sk_buff *skb, 1045 struct xfrm_state *x); 1046 void (*xdo_dev_state_advance_esn) (struct xfrm_state *x); 1047 void (*xdo_dev_state_update_stats) (struct xfrm_state *x); 1048 int (*xdo_dev_policy_add) (struct xfrm_policy *x, struct netlink_ext_ack *extack); 1049 void (*xdo_dev_policy_delete) (struct xfrm_policy *x); 1050 void (*xdo_dev_policy_free) (struct xfrm_policy *x); 1051 }; 1052 #endif 1053 1054 struct dev_ifalias { 1055 struct rcu_head rcuhead; 1056 char ifalias[]; 1057 }; 1058 1059 struct devlink; 1060 struct tlsdev_ops; 1061 1062 struct netdev_net_notifier { 1063 struct list_head list; 1064 struct notifier_block *nb; 1065 }; 1066 1067 /* 1068 * This structure defines the management hooks for network devices. 1069 * The following hooks can be defined; unless noted otherwise, they are 1070 * optional and can be filled with a null pointer. 1071 * 1072 * int (*ndo_init)(struct net_device *dev); 1073 * This function is called once when a network device is registered. 1074 * The network device can use this for any late stage initialization 1075 * or semantic validation. It can fail with an error code which will 1076 * be propagated back to register_netdev. 1077 * 1078 * void (*ndo_uninit)(struct net_device *dev); 1079 * This function is called when device is unregistered or when registration 1080 * fails. It is not called if init fails. 1081 * 1082 * int (*ndo_open)(struct net_device *dev); 1083 * This function is called when a network device transitions to the up 1084 * state. 1085 * 1086 * int (*ndo_stop)(struct net_device *dev); 1087 * This function is called when a network device transitions to the down 1088 * state. 1089 * 1090 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb, 1091 * struct net_device *dev); 1092 * Called when a packet needs to be transmitted. 1093 * Returns NETDEV_TX_OK. Can return NETDEV_TX_BUSY, but you should stop 1094 * the queue before that can happen; it's for obsolete devices and weird 1095 * corner cases, but the stack really does a non-trivial amount 1096 * of useless work if you return NETDEV_TX_BUSY. 1097 * Required; cannot be NULL. 1098 * 1099 * netdev_features_t (*ndo_features_check)(struct sk_buff *skb, 1100 * struct net_device *dev 1101 * netdev_features_t features); 1102 * Called by core transmit path to determine if device is capable of 1103 * performing offload operations on a given packet. This is to give 1104 * the device an opportunity to implement any restrictions that cannot 1105 * be otherwise expressed by feature flags. The check is called with 1106 * the set of features that the stack has calculated and it returns 1107 * those the driver believes to be appropriate. 1108 * 1109 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb, 1110 * struct net_device *sb_dev); 1111 * Called to decide which queue to use when device supports multiple 1112 * transmit queues. 1113 * 1114 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags); 1115 * This function is called to allow device receiver to make 1116 * changes to configuration when multicast or promiscuous is enabled. 1117 * 1118 * void (*ndo_set_rx_mode)(struct net_device *dev); 1119 * This function is called device changes address list filtering. 1120 * If driver handles unicast address filtering, it should set 1121 * IFF_UNICAST_FLT in its priv_flags. 1122 * Cannot sleep, called with netif_addr_lock_bh held. 1123 * Deprecated in favor of ndo_set_rx_mode_async. 1124 * 1125 * int (*ndo_set_rx_mode_async)(struct net_device *dev, 1126 * struct netdev_hw_addr_list *uc, 1127 * struct netdev_hw_addr_list *mc); 1128 * Async version of ndo_set_rx_mode which runs in process context 1129 * with rtnl_lock and netdev_lock_ops(dev) held. The uc/mc parameters 1130 * are snapshots of the address lists - iterate with 1131 * netdev_hw_addr_list_for_each(ha, uc). Return 0 on success or a 1132 * negative errno to request a retry via the core backoff. 1133 * 1134 * void (*ndo_work)(struct net_device *dev, unsigned long events); 1135 * Run deferred work scheduled with netdev_work_sched(@events). 1136 * 1137 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr); 1138 * This function is called when the Media Access Control address 1139 * needs to be changed. If this interface is not defined, the 1140 * MAC address can not be changed. 1141 * 1142 * int (*ndo_validate_addr)(struct net_device *dev); 1143 * Test if Media Access Control address is valid for the device. 1144 * 1145 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd); 1146 * Old-style ioctl entry point. This is used internally by the 1147 * ieee802154 subsystem but is no longer called by the device 1148 * ioctl handler. 1149 * 1150 * int (*ndo_siocbond)(struct net_device *dev, struct ifreq *ifr, int cmd); 1151 * Used by the bonding driver for its device specific ioctls: 1152 * SIOCBONDENSLAVE, SIOCBONDRELEASE, SIOCBONDSETHWADDR, SIOCBONDCHANGEACTIVE, 1153 * SIOCBONDSLAVEINFOQUERY, and SIOCBONDINFOQUERY 1154 * 1155 * * int (*ndo_eth_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd); 1156 * Called for ethernet specific ioctls: SIOCGMIIPHY, SIOCGMIIREG and 1157 * SIOCSMIIREG. 1158 * 1159 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map); 1160 * Used to set network devices bus interface parameters. This interface 1161 * is retained for legacy reasons; new devices should use the bus 1162 * interface (PCI) for low level management. 1163 * 1164 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu); 1165 * Called when a user wants to change the Maximum Transfer Unit 1166 * of a device. 1167 * 1168 * void (*ndo_tx_timeout)(struct net_device *dev, unsigned int txqueue); 1169 * Callback used when the transmitter has not made any progress 1170 * for dev->watchdog ticks. 1171 * 1172 * void (*ndo_get_stats64)(struct net_device *dev, 1173 * struct rtnl_link_stats64 *storage); 1174 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev); 1175 * Called when a user wants to get the network device usage 1176 * statistics. Drivers must do one of the following: 1177 * 1. Define @ndo_get_stats64 to fill in a zero-initialised 1178 * rtnl_link_stats64 structure passed by the caller. 1179 * 2. Define @ndo_get_stats to update a net_device_stats structure 1180 * (which should normally be dev->stats) and return a pointer to 1181 * it. The structure may be changed asynchronously only if each 1182 * field is written atomically. 1183 * 3. Update dev->stats asynchronously and atomically, and define 1184 * neither operation. 1185 * 1186 * bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id) 1187 * Return true if this device supports offload stats of this attr_id. 1188 * 1189 * int (*ndo_get_offload_stats)(int attr_id, const struct net_device *dev, 1190 * void *attr_data) 1191 * Get statistics for offload operations by attr_id. Write it into the 1192 * attr_data pointer. 1193 * 1194 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16 vid); 1195 * If device supports VLAN filtering this function is called when a 1196 * VLAN id is registered. 1197 * 1198 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, __be16 proto, u16 vid); 1199 * If device supports VLAN filtering this function is called when a 1200 * VLAN id is unregistered. 1201 * 1202 * void (*ndo_poll_controller)(struct net_device *dev); 1203 * 1204 * SR-IOV management functions. 1205 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac); 1206 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan, 1207 * u8 qos, __be16 proto); 1208 * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate, 1209 * int max_tx_rate); 1210 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting); 1211 * int (*ndo_set_vf_trust)(struct net_device *dev, int vf, bool setting); 1212 * int (*ndo_get_vf_config)(struct net_device *dev, 1213 * int vf, struct ifla_vf_info *ivf); 1214 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state); 1215 * int (*ndo_set_vf_port)(struct net_device *dev, int vf, 1216 * struct nlattr *port[]); 1217 * 1218 * Enable or disable the VF ability to query its RSS Redirection Table and 1219 * Hash Key. This is needed since on some devices VF share this information 1220 * with PF and querying it may introduce a theoretical security risk. 1221 * int (*ndo_set_vf_rss_query_en)(struct net_device *dev, int vf, bool setting); 1222 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb); 1223 * int (*ndo_setup_tc)(struct net_device *dev, enum tc_setup_type type, 1224 * void *type_data); 1225 * Called to setup any 'tc' scheduler, classifier or action on @dev. 1226 * This is always called from the stack with the rtnl lock held and netif 1227 * tx queues stopped. This allows the netdevice to perform queue 1228 * management safely. 1229 * 1230 * NB: Returning -EOPNOTSUPP for whatever commands means "this qdisc 1231 * is not offloaded (anymore, offloading may have silently stopped)", 1232 * and the offloading flag is cleared. Notably, this is also true for 1233 * dump queries (e.g. TC_*_STATS commands). If the underlying device does 1234 * not report any statistics but is still offloading, return 0 instead. 1235 * 1236 * Fiber Channel over Ethernet (FCoE) offload functions. 1237 * int (*ndo_fcoe_enable)(struct net_device *dev); 1238 * Called when the FCoE protocol stack wants to start using LLD for FCoE 1239 * so the underlying device can perform whatever needed configuration or 1240 * initialization to support acceleration of FCoE traffic. 1241 * 1242 * int (*ndo_fcoe_disable)(struct net_device *dev); 1243 * Called when the FCoE protocol stack wants to stop using LLD for FCoE 1244 * so the underlying device can perform whatever needed clean-ups to 1245 * stop supporting acceleration of FCoE traffic. 1246 * 1247 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid, 1248 * struct scatterlist *sgl, unsigned int sgc); 1249 * Called when the FCoE Initiator wants to initialize an I/O that 1250 * is a possible candidate for Direct Data Placement (DDP). The LLD can 1251 * perform necessary setup and returns 1 to indicate the device is set up 1252 * successfully to perform DDP on this I/O, otherwise this returns 0. 1253 * 1254 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid); 1255 * Called when the FCoE Initiator/Target is done with the DDPed I/O as 1256 * indicated by the FC exchange id 'xid', so the underlying device can 1257 * clean up and reuse resources for later DDP requests. 1258 * 1259 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid, 1260 * struct scatterlist *sgl, unsigned int sgc); 1261 * Called when the FCoE Target wants to initialize an I/O that 1262 * is a possible candidate for Direct Data Placement (DDP). The LLD can 1263 * perform necessary setup and returns 1 to indicate the device is set up 1264 * successfully to perform DDP on this I/O, otherwise this returns 0. 1265 * 1266 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev, 1267 * struct netdev_fcoe_hbainfo *hbainfo); 1268 * Called when the FCoE Protocol stack wants information on the underlying 1269 * device. This information is utilized by the FCoE protocol stack to 1270 * register attributes with Fiber Channel management service as per the 1271 * FC-GS Fabric Device Management Information(FDMI) specification. 1272 * 1273 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type); 1274 * Called when the underlying device wants to override default World Wide 1275 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own 1276 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE 1277 * protocol stack to use. 1278 * 1279 * RFS acceleration. 1280 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb, 1281 * u16 rxq_index, u32 flow_id); 1282 * Set hardware filter for RFS. rxq_index is the target queue index; 1283 * flow_id is a flow ID to be passed to rps_may_expire_flow() later. 1284 * Return the filter ID on success, or a negative error code. 1285 * 1286 * Slave management functions (for bridge, bonding, etc). 1287 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev); 1288 * Called to make another netdev an underling. 1289 * 1290 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev); 1291 * Called to release previously enslaved netdev. 1292 * 1293 * struct net_device *(*ndo_get_xmit_slave)(struct net_device *dev, 1294 * struct sk_buff *skb, 1295 * bool all_slaves); 1296 * Get the xmit slave of master device. If all_slaves is true, function 1297 * assume all the slaves can transmit. 1298 * 1299 * Feature/offload setting functions. 1300 * netdev_features_t (*ndo_fix_features)(struct net_device *dev, 1301 * netdev_features_t features); 1302 * Adjusts the requested feature flags according to device-specific 1303 * constraints, and returns the resulting flags. Must not modify 1304 * the device state. 1305 * 1306 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features); 1307 * Called to update device configuration to new features. Passed 1308 * feature set might be less than what was returned by ndo_fix_features()). 1309 * Must return >0 or -errno if it changed dev->features itself. 1310 * 1311 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[], 1312 * struct net_device *dev, 1313 * const unsigned char *addr, u16 vid, u16 flags, 1314 * bool *notified, struct netlink_ext_ack *extack); 1315 * Adds an FDB entry to dev for addr. 1316 * Callee shall set *notified to true if it sent any appropriate 1317 * notification(s). Otherwise core will send a generic one. 1318 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[], 1319 * struct net_device *dev, 1320 * const unsigned char *addr, u16 vid 1321 * bool *notified, struct netlink_ext_ack *extack); 1322 * Deletes the FDB entry from dev corresponding to addr. 1323 * Callee shall set *notified to true if it sent any appropriate 1324 * notification(s). Otherwise core will send a generic one. 1325 * int (*ndo_fdb_del_bulk)(struct nlmsghdr *nlh, struct net_device *dev, 1326 * struct netlink_ext_ack *extack); 1327 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb, 1328 * struct net_device *dev, struct net_device *filter_dev, 1329 * int *idx) 1330 * Used to add FDB entries to dump requests. Implementers should add 1331 * entries to skb and update idx with the number of entries. 1332 * 1333 * int (*ndo_mdb_add)(struct net_device *dev, struct nlattr *tb[], 1334 * u16 nlmsg_flags, struct netlink_ext_ack *extack); 1335 * Adds an MDB entry to dev. 1336 * int (*ndo_mdb_del)(struct net_device *dev, struct nlattr *tb[], 1337 * struct netlink_ext_ack *extack); 1338 * Deletes the MDB entry from dev. 1339 * int (*ndo_mdb_del_bulk)(struct net_device *dev, struct nlattr *tb[], 1340 * struct netlink_ext_ack *extack); 1341 * Bulk deletes MDB entries from dev. 1342 * int (*ndo_mdb_dump)(struct net_device *dev, struct sk_buff *skb, 1343 * struct netlink_callback *cb); 1344 * Dumps MDB entries from dev. The first argument (marker) in the netlink 1345 * callback is used by core rtnetlink code. 1346 * 1347 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh, 1348 * u16 flags, struct netlink_ext_ack *extack) 1349 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq, 1350 * struct net_device *dev, u32 filter_mask, 1351 * int nlflags) 1352 * int (*ndo_bridge_dellink)(struct net_device *dev, struct nlmsghdr *nlh, 1353 * u16 flags); 1354 * 1355 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier); 1356 * Called to change device carrier. Soft-devices (like dummy, team, etc) 1357 * which do not represent real hardware may define this to allow their 1358 * userspace components to manage their virtual carrier state. Devices 1359 * that determine carrier state from physical hardware properties (eg 1360 * network cables) or protocol-dependent mechanisms (eg 1361 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function. 1362 * 1363 * int (*ndo_get_phys_port_id)(struct net_device *dev, 1364 * struct netdev_phys_item_id *ppid); 1365 * Called to get ID of physical port of this device. If driver does 1366 * not implement this, it is assumed that the hw is not able to have 1367 * multiple net devices on single physical port. 1368 * 1369 * int (*ndo_get_port_parent_id)(struct net_device *dev, 1370 * struct netdev_phys_item_id *ppid) 1371 * Called to get the parent ID of the physical port of this device. 1372 * 1373 * void* (*ndo_dfwd_add_station)(struct net_device *pdev, 1374 * struct net_device *dev) 1375 * Called by upper layer devices to accelerate switching or other 1376 * station functionality into hardware. 'pdev is the lowerdev 1377 * to use for the offload and 'dev' is the net device that will 1378 * back the offload. Returns a pointer to the private structure 1379 * the upper layer will maintain. 1380 * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv) 1381 * Called by upper layer device to delete the station created 1382 * by 'ndo_dfwd_add_station'. 'pdev' is the net device backing 1383 * the station and priv is the structure returned by the add 1384 * operation. 1385 * int (*ndo_set_tx_maxrate)(struct net_device *dev, 1386 * int queue_index, u32 maxrate); 1387 * Called when a user wants to set a max-rate limitation of specific 1388 * TX queue. 1389 * int (*ndo_get_iflink)(const struct net_device *dev); 1390 * Called to get the iflink value of this device. 1391 * int (*ndo_fill_metadata_dst)(struct net_device *dev, struct sk_buff *skb); 1392 * This function is used to get egress tunnel information for given skb. 1393 * This is useful for retrieving outer tunnel header parameters while 1394 * sampling packet. 1395 * void (*ndo_set_rx_headroom)(struct net_device *dev, int needed_headroom); 1396 * This function is used to specify the headroom that the skb must 1397 * consider when allocation skb during packet reception. Setting 1398 * appropriate rx headroom value allows avoiding skb head copy on 1399 * forward. Setting a negative value resets the rx headroom to the 1400 * default value. 1401 * int (*ndo_bpf)(struct net_device *dev, struct netdev_bpf *bpf); 1402 * This function is used to set or query state related to XDP on the 1403 * netdevice and manage BPF offload. See definition of 1404 * enum bpf_netdev_command for details. 1405 * int (*ndo_xdp_xmit)(struct net_device *dev, int n, struct xdp_frame **xdp, 1406 * u32 flags); 1407 * This function is used to submit @n XDP packets for transmit on a 1408 * netdevice. Returns number of frames successfully transmitted, frames 1409 * that got dropped are freed/returned via xdp_return_frame(). 1410 * Returns negative number, means general error invoking ndo, meaning 1411 * no frames were xmit'ed and core-caller will free all frames. 1412 * struct net_device *(*ndo_xdp_get_xmit_slave)(struct net_device *dev, 1413 * struct xdp_buff *xdp); 1414 * Get the xmit slave of master device based on the xdp_buff. 1415 * int (*ndo_xsk_wakeup)(struct net_device *dev, u32 queue_id, u32 flags); 1416 * This function is used to wake up the softirq, ksoftirqd or kthread 1417 * responsible for sending and/or receiving packets on a specific 1418 * queue id bound to an AF_XDP socket. The flags field specifies if 1419 * only RX, only Tx, or both should be woken up using the flags 1420 * XDP_WAKEUP_RX and XDP_WAKEUP_TX. 1421 * int (*ndo_tunnel_ctl)(struct net_device *dev, struct ip_tunnel_parm_kern *p, 1422 * int cmd); 1423 * Add, change, delete or get information on an IPv4 tunnel. 1424 * struct net_device *(*ndo_get_peer_dev)(struct net_device *dev); 1425 * If a device is paired with a peer device, return the peer instance. 1426 * The caller must be under RCU read context. 1427 * int (*ndo_fill_forward_path)(struct net_device_path_ctx *ctx, struct net_device_path *path); 1428 * Get the forwarding path to reach the real device from the HW destination address 1429 * ktime_t (*ndo_get_tstamp)(struct net_device *dev, 1430 * const struct skb_shared_hwtstamps *hwtstamps, 1431 * bool cycles); 1432 * Get hardware timestamp based on normal/adjustable time or free running 1433 * cycle counter. This function is required if physical clock supports a 1434 * free running cycle counter. 1435 * 1436 * int (*ndo_hwtstamp_get)(struct net_device *dev, 1437 * struct kernel_hwtstamp_config *kernel_config); 1438 * Get the currently configured hardware timestamping parameters for the 1439 * NIC device. 1440 * 1441 * int (*ndo_hwtstamp_set)(struct net_device *dev, 1442 * struct kernel_hwtstamp_config *kernel_config, 1443 * struct netlink_ext_ack *extack); 1444 * Change the hardware timestamping parameters for NIC device. 1445 */ 1446 struct net_device_ops { 1447 int (*ndo_init)(struct net_device *dev); 1448 void (*ndo_uninit)(struct net_device *dev); 1449 int (*ndo_open)(struct net_device *dev); 1450 int (*ndo_stop)(struct net_device *dev); 1451 netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb, 1452 struct net_device *dev); 1453 netdev_features_t (*ndo_features_check)(struct sk_buff *skb, 1454 struct net_device *dev, 1455 netdev_features_t features); 1456 u16 (*ndo_select_queue)(struct net_device *dev, 1457 struct sk_buff *skb, 1458 struct net_device *sb_dev); 1459 void (*ndo_change_rx_flags)(struct net_device *dev, 1460 int flags); 1461 void (*ndo_set_rx_mode)(struct net_device *dev); 1462 int (*ndo_set_rx_mode_async)( 1463 struct net_device *dev, 1464 struct netdev_hw_addr_list *uc, 1465 struct netdev_hw_addr_list *mc); 1466 void (*ndo_work)(struct net_device *dev, 1467 unsigned long events); 1468 int (*ndo_set_mac_address)(struct net_device *dev, 1469 void *addr); 1470 int (*ndo_validate_addr)(struct net_device *dev); 1471 int (*ndo_do_ioctl)(struct net_device *dev, 1472 struct ifreq *ifr, int cmd); 1473 int (*ndo_eth_ioctl)(struct net_device *dev, 1474 struct ifreq *ifr, int cmd); 1475 int (*ndo_siocbond)(struct net_device *dev, 1476 struct ifreq *ifr, int cmd); 1477 int (*ndo_siocwandev)(struct net_device *dev, 1478 struct if_settings *ifs); 1479 int (*ndo_siocdevprivate)(struct net_device *dev, 1480 struct ifreq *ifr, 1481 void __user *data, int cmd); 1482 int (*ndo_set_config)(struct net_device *dev, 1483 struct ifmap *map); 1484 int (*ndo_change_mtu)(struct net_device *dev, 1485 int new_mtu); 1486 int (*ndo_neigh_setup)(struct net_device *dev, 1487 struct neigh_parms *); 1488 void (*ndo_tx_timeout) (struct net_device *dev, 1489 unsigned int txqueue); 1490 1491 void (*ndo_get_stats64)(struct net_device *dev, 1492 struct rtnl_link_stats64 *storage); 1493 bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id); 1494 int (*ndo_get_offload_stats)(int attr_id, 1495 const struct net_device *dev, 1496 void *attr_data); 1497 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev); 1498 1499 int (*ndo_vlan_rx_add_vid)(struct net_device *dev, 1500 __be16 proto, u16 vid); 1501 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, 1502 __be16 proto, u16 vid); 1503 #ifdef CONFIG_NET_POLL_CONTROLLER 1504 void (*ndo_poll_controller)(struct net_device *dev); 1505 int (*ndo_netpoll_setup)(struct net_device *dev); 1506 void (*ndo_netpoll_cleanup)(struct net_device *dev); 1507 #endif 1508 int (*ndo_set_vf_mac)(struct net_device *dev, 1509 int queue, u8 *mac); 1510 int (*ndo_set_vf_vlan)(struct net_device *dev, 1511 int queue, u16 vlan, 1512 u8 qos, __be16 proto); 1513 int (*ndo_set_vf_rate)(struct net_device *dev, 1514 int vf, int min_tx_rate, 1515 int max_tx_rate); 1516 int (*ndo_set_vf_spoofchk)(struct net_device *dev, 1517 int vf, bool setting); 1518 int (*ndo_set_vf_trust)(struct net_device *dev, 1519 int vf, bool setting); 1520 int (*ndo_get_vf_config)(struct net_device *dev, 1521 int vf, 1522 struct ifla_vf_info *ivf); 1523 int (*ndo_set_vf_link_state)(struct net_device *dev, 1524 int vf, int link_state); 1525 int (*ndo_get_vf_stats)(struct net_device *dev, 1526 int vf, 1527 struct ifla_vf_stats 1528 *vf_stats); 1529 int (*ndo_set_vf_port)(struct net_device *dev, 1530 int vf, 1531 struct nlattr *port[]); 1532 int (*ndo_get_vf_port)(struct net_device *dev, 1533 int vf, struct sk_buff *skb); 1534 int (*ndo_get_vf_guid)(struct net_device *dev, 1535 int vf, 1536 struct ifla_vf_guid *node_guid, 1537 struct ifla_vf_guid *port_guid); 1538 int (*ndo_set_vf_guid)(struct net_device *dev, 1539 int vf, u64 guid, 1540 int guid_type); 1541 int (*ndo_set_vf_rss_query_en)( 1542 struct net_device *dev, 1543 int vf, bool setting); 1544 int (*ndo_setup_tc)(struct net_device *dev, 1545 enum tc_setup_type type, 1546 void *type_data); 1547 #if IS_ENABLED(CONFIG_FCOE) 1548 int (*ndo_fcoe_enable)(struct net_device *dev); 1549 int (*ndo_fcoe_disable)(struct net_device *dev); 1550 int (*ndo_fcoe_ddp_setup)(struct net_device *dev, 1551 u16 xid, 1552 struct scatterlist *sgl, 1553 unsigned int sgc); 1554 int (*ndo_fcoe_ddp_done)(struct net_device *dev, 1555 u16 xid); 1556 int (*ndo_fcoe_ddp_target)(struct net_device *dev, 1557 u16 xid, 1558 struct scatterlist *sgl, 1559 unsigned int sgc); 1560 int (*ndo_fcoe_get_hbainfo)(struct net_device *dev, 1561 struct netdev_fcoe_hbainfo *hbainfo); 1562 #endif 1563 1564 #if IS_ENABLED(CONFIG_LIBFCOE) 1565 #define NETDEV_FCOE_WWNN 0 1566 #define NETDEV_FCOE_WWPN 1 1567 int (*ndo_fcoe_get_wwn)(struct net_device *dev, 1568 u64 *wwn, int type); 1569 #endif 1570 1571 #ifdef CONFIG_RFS_ACCEL 1572 int (*ndo_rx_flow_steer)(struct net_device *dev, 1573 const struct sk_buff *skb, 1574 u16 rxq_index, 1575 u32 flow_id); 1576 #endif 1577 int (*ndo_add_slave)(struct net_device *dev, 1578 struct net_device *slave_dev, 1579 struct netlink_ext_ack *extack); 1580 int (*ndo_del_slave)(struct net_device *dev, 1581 struct net_device *slave_dev); 1582 struct net_device* (*ndo_get_xmit_slave)(struct net_device *dev, 1583 struct sk_buff *skb, 1584 bool all_slaves); 1585 struct net_device* (*ndo_sk_get_lower_dev)(struct net_device *dev, 1586 struct sock *sk); 1587 netdev_features_t (*ndo_fix_features)(struct net_device *dev, 1588 netdev_features_t features); 1589 int (*ndo_set_features)(struct net_device *dev, 1590 netdev_features_t features); 1591 int (*ndo_neigh_construct)(struct net_device *dev, 1592 struct neighbour *n); 1593 void (*ndo_neigh_destroy)(struct net_device *dev, 1594 struct neighbour *n); 1595 1596 int (*ndo_fdb_add)(struct ndmsg *ndm, 1597 struct nlattr *tb[], 1598 struct net_device *dev, 1599 const unsigned char *addr, 1600 u16 vid, 1601 u16 flags, 1602 bool *notified, 1603 struct netlink_ext_ack *extack); 1604 int (*ndo_fdb_del)(struct ndmsg *ndm, 1605 struct nlattr *tb[], 1606 struct net_device *dev, 1607 const unsigned char *addr, 1608 u16 vid, 1609 bool *notified, 1610 struct netlink_ext_ack *extack); 1611 int (*ndo_fdb_del_bulk)(struct nlmsghdr *nlh, 1612 struct net_device *dev, 1613 struct netlink_ext_ack *extack); 1614 int (*ndo_fdb_dump)(struct sk_buff *skb, 1615 struct netlink_callback *cb, 1616 struct net_device *dev, 1617 struct net_device *filter_dev, 1618 int *idx); 1619 int (*ndo_fdb_get)(struct sk_buff *skb, 1620 struct nlattr *tb[], 1621 struct net_device *dev, 1622 const unsigned char *addr, 1623 u16 vid, u32 portid, u32 seq, 1624 struct netlink_ext_ack *extack); 1625 int (*ndo_mdb_add)(struct net_device *dev, 1626 struct nlattr *tb[], 1627 u16 nlmsg_flags, 1628 struct netlink_ext_ack *extack); 1629 int (*ndo_mdb_del)(struct net_device *dev, 1630 struct nlattr *tb[], 1631 struct netlink_ext_ack *extack); 1632 int (*ndo_mdb_del_bulk)(struct net_device *dev, 1633 struct nlattr *tb[], 1634 struct netlink_ext_ack *extack); 1635 int (*ndo_mdb_dump)(struct net_device *dev, 1636 struct sk_buff *skb, 1637 struct netlink_callback *cb); 1638 int (*ndo_mdb_get)(struct net_device *dev, 1639 struct nlattr *tb[], u32 portid, 1640 u32 seq, 1641 struct netlink_ext_ack *extack); 1642 int (*ndo_bridge_setlink)(struct net_device *dev, 1643 struct nlmsghdr *nlh, 1644 u16 flags, 1645 struct netlink_ext_ack *extack); 1646 int (*ndo_bridge_getlink)(struct sk_buff *skb, 1647 u32 pid, u32 seq, 1648 struct net_device *dev, 1649 u32 filter_mask, 1650 int nlflags); 1651 int (*ndo_bridge_dellink)(struct net_device *dev, 1652 struct nlmsghdr *nlh, 1653 u16 flags); 1654 int (*ndo_change_carrier)(struct net_device *dev, 1655 bool new_carrier); 1656 int (*ndo_get_phys_port_id)(struct net_device *dev, 1657 struct netdev_phys_item_id *ppid); 1658 int (*ndo_get_port_parent_id)(struct net_device *dev, 1659 struct netdev_phys_item_id *ppid); 1660 int (*ndo_get_phys_port_name)(struct net_device *dev, 1661 char *name, size_t len); 1662 void* (*ndo_dfwd_add_station)(struct net_device *pdev, 1663 struct net_device *dev); 1664 void (*ndo_dfwd_del_station)(struct net_device *pdev, 1665 void *priv); 1666 1667 int (*ndo_set_tx_maxrate)(struct net_device *dev, 1668 int queue_index, 1669 u32 maxrate); 1670 int (*ndo_get_iflink)(const struct net_device *dev); 1671 int (*ndo_fill_metadata_dst)(struct net_device *dev, 1672 struct sk_buff *skb); 1673 void (*ndo_set_rx_headroom)(struct net_device *dev, 1674 int needed_headroom); 1675 int (*ndo_bpf)(struct net_device *dev, 1676 struct netdev_bpf *bpf); 1677 int (*ndo_xdp_xmit)(struct net_device *dev, int n, 1678 struct xdp_frame **xdp, 1679 u32 flags); 1680 struct net_device * (*ndo_xdp_get_xmit_slave)(struct net_device *dev, 1681 struct xdp_buff *xdp); 1682 int (*ndo_xsk_wakeup)(struct net_device *dev, 1683 u32 queue_id, u32 flags); 1684 int (*ndo_tunnel_ctl)(struct net_device *dev, 1685 struct ip_tunnel_parm_kern *p, 1686 int cmd); 1687 struct net_device * (*ndo_get_peer_dev)(struct net_device *dev); 1688 int (*ndo_fill_forward_path)(struct net_device_path_ctx *ctx, 1689 struct net_device_path *path); 1690 ktime_t (*ndo_get_tstamp)(struct net_device *dev, 1691 const struct skb_shared_hwtstamps *hwtstamps, 1692 bool cycles); 1693 int (*ndo_hwtstamp_get)(struct net_device *dev, 1694 struct kernel_hwtstamp_config *kernel_config); 1695 int (*ndo_hwtstamp_set)(struct net_device *dev, 1696 struct kernel_hwtstamp_config *kernel_config, 1697 struct netlink_ext_ack *extack); 1698 1699 #if IS_ENABLED(CONFIG_NET_SHAPER) 1700 /** 1701 * @net_shaper_ops: Device shaping offload operations 1702 * see include/net/net_shapers.h 1703 */ 1704 const struct net_shaper_ops *net_shaper_ops; 1705 #endif 1706 }; 1707 1708 /** 1709 * enum netdev_priv_flags - &struct net_device priv_flags 1710 * 1711 * These are the &struct net_device, they are only set internally 1712 * by drivers and used in the kernel. These flags are invisible to 1713 * userspace; this means that the order of these flags can change 1714 * during any kernel release. 1715 * 1716 * You should add bitfield booleans after either net_device::priv_flags 1717 * (hotpath) or ::threaded (slowpath) instead of extending these flags. 1718 * 1719 * @IFF_802_1Q_VLAN: 802.1Q VLAN device 1720 * @IFF_EBRIDGE: Ethernet bridging device 1721 * @IFF_BONDING: bonding master or slave 1722 * @IFF_ISATAP: ISATAP interface (RFC4214) 1723 * @IFF_WAN_HDLC: WAN HDLC device 1724 * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to 1725 * release skb->dst 1726 * @IFF_DONT_BRIDGE: disallow bridging this ether dev 1727 * @IFF_DISABLE_NETPOLL: disable netpoll at run-time 1728 * @IFF_MACVLAN_PORT: device used as macvlan port 1729 * @IFF_BRIDGE_PORT: device used as bridge port 1730 * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port 1731 * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit 1732 * @IFF_UNICAST_FLT: Supports unicast filtering 1733 * @IFF_TEAM_PORT: device used as team port 1734 * @IFF_SUPP_NOFCS: device supports sending custom FCS 1735 * @IFF_LIVE_ADDR_CHANGE: device supports hardware address 1736 * change when it's running 1737 * @IFF_MACVLAN: Macvlan device 1738 * @IFF_XMIT_DST_RELEASE_PERM: IFF_XMIT_DST_RELEASE not taking into account 1739 * underlying stacked devices 1740 * @IFF_L3MDEV_MASTER: device is an L3 master device 1741 * @IFF_NO_QUEUE: device can run without qdisc attached 1742 * @IFF_OPENVSWITCH: device is a Open vSwitch master 1743 * @IFF_L3MDEV_SLAVE: device is enslaved to an L3 master device 1744 * @IFF_TEAM: device is a team device 1745 * @IFF_PHONY_HEADROOM: the headroom value is controlled by an external 1746 * entity (i.e. the master device for bridged veth) 1747 * @IFF_MACSEC: device is a MACsec device 1748 * @IFF_NO_RX_HANDLER: device doesn't support the rx_handler hook 1749 * @IFF_FAILOVER: device is a failover master device 1750 * @IFF_FAILOVER_SLAVE: device is lower dev of a failover master device 1751 * @IFF_L3MDEV_RX_HANDLER: only invoke the rx handler of L3 master device 1752 * @IFF_NO_ADDRCONF: prevent ipv6 addrconf 1753 * @IFF_TX_SKB_NO_LINEAR: device/driver is capable of xmitting frames with 1754 * skb_headlen(skb) == 0 (data starts from frag0) 1755 */ 1756 enum netdev_priv_flags { 1757 IFF_802_1Q_VLAN = 1<<0, 1758 IFF_EBRIDGE = 1<<1, 1759 IFF_BONDING = 1<<2, 1760 IFF_ISATAP = 1<<3, 1761 IFF_WAN_HDLC = 1<<4, 1762 IFF_XMIT_DST_RELEASE = 1<<5, 1763 IFF_DONT_BRIDGE = 1<<6, 1764 IFF_DISABLE_NETPOLL = 1<<7, 1765 IFF_MACVLAN_PORT = 1<<8, 1766 IFF_BRIDGE_PORT = 1<<9, 1767 IFF_OVS_DATAPATH = 1<<10, 1768 IFF_TX_SKB_SHARING = 1<<11, 1769 IFF_UNICAST_FLT = 1<<12, 1770 IFF_TEAM_PORT = 1<<13, 1771 IFF_SUPP_NOFCS = 1<<14, 1772 IFF_LIVE_ADDR_CHANGE = 1<<15, 1773 IFF_MACVLAN = 1<<16, 1774 IFF_XMIT_DST_RELEASE_PERM = 1<<17, 1775 IFF_L3MDEV_MASTER = 1<<18, 1776 IFF_NO_QUEUE = 1<<19, 1777 IFF_OPENVSWITCH = 1<<20, 1778 IFF_L3MDEV_SLAVE = 1<<21, 1779 IFF_TEAM = 1<<22, 1780 IFF_PHONY_HEADROOM = 1<<24, 1781 IFF_MACSEC = 1<<25, 1782 IFF_NO_RX_HANDLER = 1<<26, 1783 IFF_FAILOVER = 1<<27, 1784 IFF_FAILOVER_SLAVE = 1<<28, 1785 IFF_L3MDEV_RX_HANDLER = 1<<29, 1786 IFF_NO_ADDRCONF = BIT_ULL(30), 1787 IFF_TX_SKB_NO_LINEAR = BIT_ULL(31), 1788 }; 1789 1790 /* Specifies the type of the struct net_device::ml_priv pointer */ 1791 enum netdev_ml_priv_type { 1792 ML_PRIV_NONE, 1793 ML_PRIV_CAN, 1794 }; 1795 1796 enum netdev_stat_type { 1797 NETDEV_PCPU_STAT_NONE, 1798 NETDEV_PCPU_STAT_LSTATS, /* struct pcpu_lstats */ 1799 NETDEV_PCPU_STAT_TSTATS, /* struct pcpu_sw_netstats */ 1800 NETDEV_PCPU_STAT_DSTATS, /* struct pcpu_dstats */ 1801 }; 1802 1803 enum netmem_tx_mode { 1804 NETMEM_TX_NONE, /* no netmem TX support */ 1805 NETMEM_TX_DMA, /* DMA-capable netmem TX (real HW) */ 1806 NETMEM_TX_NO_DMA, /* no DMA, e.g. passthrough for virtual devs */ 1807 }; 1808 1809 enum netdev_reg_state { 1810 NETREG_UNINITIALIZED = 0, 1811 NETREG_REGISTERED, /* completed register_netdevice */ 1812 NETREG_UNREGISTERING, /* called unregister_netdevice */ 1813 NETREG_UNREGISTERED, /* completed unregister todo */ 1814 NETREG_RELEASED, /* called free_netdev */ 1815 NETREG_DUMMY, /* dummy device for NAPI poll */ 1816 }; 1817 1818 /** 1819 * struct net_device - The DEVICE structure. 1820 * 1821 * Actually, this whole structure is a big mistake. It mixes I/O 1822 * data with strictly "high-level" data, and it has to know about 1823 * almost every data structure used in the INET module. 1824 * 1825 * @priv_flags: flags invisible to userspace defined as bits, see 1826 * enum netdev_priv_flags for the definitions 1827 * @lltx: device supports lockless Tx. Deprecated for real HW 1828 * drivers. Mainly used by logical interfaces, such as 1829 * bonding and tunnels 1830 * @netmem_tx: device netmem TX mode 1831 * 1832 * @name: This is the first field of the "visible" part of this structure 1833 * (i.e. as seen by users in the "Space.c" file). It is the name 1834 * of the interface. 1835 * 1836 * @name_node: Name hashlist node 1837 * @ifalias: SNMP alias 1838 * @mem_end: Shared memory end 1839 * @mem_start: Shared memory start 1840 * @base_addr: Device I/O address 1841 * @irq: Device IRQ number 1842 * 1843 * @state: Generic network queuing layer state, see netdev_state_t 1844 * @dev_list: The global list of network devices 1845 * @napi_list: List entry used for polling NAPI devices 1846 * @unreg_list: List entry when we are unregistering the 1847 * device; see the function unregister_netdev 1848 * @close_list: List entry used when we are closing the device 1849 * @ptype_all: Device-specific packet handlers for all protocols 1850 * @ptype_specific: Device-specific, protocol-specific packet handlers 1851 * 1852 * @adj_list: Directly linked devices, like slaves for bonding 1853 * @features: Currently active device features 1854 * @hw_features: User-changeable features 1855 * 1856 * @wanted_features: User-requested features 1857 * @vlan_features: Mask of features inheritable by VLAN devices 1858 * 1859 * @hw_enc_features: Mask of features inherited by encapsulating devices 1860 * This field indicates what encapsulation 1861 * offloads the hardware is capable of doing, 1862 * and drivers will need to set them appropriately. 1863 * 1864 * @mpls_features: Mask of features inheritable by MPLS 1865 * @gso_partial_features: value(s) from NETIF_F_GSO\* 1866 * @mangleid_features: Mask of features requiring MANGLEID, will be 1867 * disabled together with the latter. 1868 * 1869 * @ifindex: interface index 1870 * @group: The group the device belongs to 1871 * 1872 * @stats: Statistics struct, which was left as a legacy, use 1873 * rtnl_link_stats64 instead 1874 * 1875 * @core_stats: core networking counters, 1876 * do not use this in drivers 1877 * @carrier_up_count: Number of times the carrier has been up 1878 * @carrier_down_count: Number of times the carrier has been down 1879 * 1880 * @wireless_handlers: List of functions to handle Wireless Extensions, 1881 * instead of ioctl, 1882 * see <net/iw_handler.h> for details. 1883 * 1884 * @netdev_ops: Includes several pointers to callbacks, 1885 * if one wants to override the ndo_*() functions 1886 * @xdp_metadata_ops: Includes pointers to XDP metadata callbacks. 1887 * @xsk_tx_metadata_ops: Includes pointers to AF_XDP TX metadata callbacks. 1888 * @ethtool_ops: Management operations 1889 * @l3mdev_ops: Layer 3 master device operations 1890 * @ndisc_ops: Includes callbacks for different IPv6 neighbour 1891 * discovery handling. Necessary for e.g. 6LoWPAN. 1892 * @xfrmdev_ops: Transformation offload operations 1893 * @tlsdev_ops: Transport Layer Security offload operations 1894 * @header_ops: Includes callbacks for creating,parsing,caching,etc 1895 * of Layer 2 headers. 1896 * 1897 * @flags: Interface flags (a la BSD) 1898 * @xdp_features: XDP capability supported by the device 1899 * @gflags: Global flags ( kept as legacy ) 1900 * @priv_len: Size of the ->priv flexible array 1901 * @priv: Flexible array containing private data 1902 * @operstate: RFC2863 operstate 1903 * @link_mode: Mapping policy to operstate 1904 * @if_port: Selectable AUI, TP, ... 1905 * @dma: DMA channel 1906 * @mtu: Interface MTU value 1907 * @min_mtu: Interface Minimum MTU value 1908 * @max_mtu: Interface Maximum MTU value 1909 * @type: Interface hardware type 1910 * @hard_header_len: Maximum hardware header length. 1911 * @min_header_len: Minimum hardware header length 1912 * 1913 * @needed_headroom: Extra headroom the hardware may need, but not in all 1914 * cases can this be guaranteed 1915 * @needed_tailroom: Extra tailroom the hardware may need, but not in all 1916 * cases can this be guaranteed. Some cases also use 1917 * LL_MAX_HEADER instead to allocate the skb 1918 * 1919 * interface address info: 1920 * 1921 * @perm_addr: Permanent hw address 1922 * @addr_assign_type: Hw address assignment type 1923 * @addr_len: Hardware address length 1924 * @upper_level: Maximum depth level of upper devices. 1925 * @lower_level: Maximum depth level of lower devices. 1926 * @threaded: napi threaded state. 1927 * @neigh_priv_len: Used in neigh_alloc() 1928 * @dev_id: Used to differentiate devices that share 1929 * the same link layer address 1930 * @dev_port: Used to differentiate devices that share 1931 * the same function 1932 * @addr_list_lock: XXX: need comments on this one 1933 * @name_assign_type: network interface name assignment type 1934 * @uc_promisc: Counter that indicates promiscuous mode 1935 * has been enabled due to the need to listen to 1936 * additional unicast addresses in a device that 1937 * does not implement ndo_set_rx_mode() 1938 * @work_node: List entry for async netdev_work processing 1939 * @work_tracker: Refcount tracker for async netdev_work 1940 * @work_pending: Driver-defined pending netdev_work, passed to 1941 * ndo_work() (see netdev_work_sched()) 1942 * @work_core_pending: Core-defined pending netdev_work (NETDEV_WORK_*) 1943 * @rx_mode_addr_cache: Recycled snapshot entries for rx_mode work 1944 * @rx_mode_retry_timer: Timer that re-queues rx_mode work after failure 1945 * @rx_mode_retry_count: Number of consecutive retries already scheduled 1946 * @uc: unicast mac addresses 1947 * @mc: multicast mac addresses 1948 * @dev_addrs: list of device hw addresses 1949 * @queues_kset: Group of all Kobjects in the Tx and RX queues 1950 * @promiscuity: Number of times the NIC is told to work in 1951 * promiscuous mode; if it becomes 0 the NIC will 1952 * exit promiscuous mode 1953 * @allmulti: Counter, enables or disables allmulticast mode 1954 * 1955 * @vlan_info: VLAN info 1956 * @dsa_ptr: dsa specific data 1957 * @tipc_ptr: TIPC specific data 1958 * @ip_ptr: IPv4 specific data 1959 * @ip6_ptr: IPv6 specific data 1960 * @ieee80211_ptr: IEEE 802.11 specific data, assign before registering 1961 * @ieee802154_ptr: IEEE 802.15.4 low-rate Wireless Personal Area Network 1962 * device struct 1963 * @mpls_ptr: mpls_dev struct pointer 1964 * @mctp_ptr: MCTP specific data 1965 * @psp_dev: PSP crypto device registered for this netdev 1966 * 1967 * @dev_addr: Hw address (before bcast, 1968 * because most packets are unicast) 1969 * 1970 * @_rx: Array of RX queues 1971 * @num_rx_queues: Number of RX queues 1972 * allocated at register_netdev() time 1973 * @real_num_rx_queues: Number of RX queues currently active in device 1974 * @xdp_prog: XDP sockets filter program pointer 1975 * 1976 * @rx_handler: handler for received packets 1977 * @rx_handler_data: XXX: need comments on this one 1978 * @tcx_ingress: BPF & clsact qdisc specific data for ingress processing 1979 * @ingress_queue: XXX: need comments on this one 1980 * @nf_hooks_ingress: netfilter hooks executed for ingress packets 1981 * @broadcast: hw bcast address 1982 * 1983 * @rx_cpu_rmap: CPU reverse-mapping for RX completion interrupts, 1984 * indexed by RX queue number. Assigned by driver. 1985 * This must only be set if the ndo_rx_flow_steer 1986 * operation is defined 1987 * @index_hlist: Device index hash chain 1988 * 1989 * @_tx: Array of TX queues 1990 * @num_tx_queues: Number of TX queues allocated at alloc_netdev_mq() time 1991 * @real_num_tx_queues: Number of TX queues currently active in device 1992 * @qdisc: Root qdisc from userspace point of view 1993 * @tx_queue_len: Max frames per queue allowed 1994 * @tx_global_lock: XXX: need comments on this one 1995 * @xdp_bulkq: XDP device bulk queue 1996 * @xps_maps: all CPUs/RXQs maps for XPS device 1997 * 1998 * @xps_maps: XXX: need comments on this one 1999 * @tcx_egress: BPF & clsact qdisc specific data for egress processing 2000 * @nf_hooks_egress: netfilter hooks executed for egress packets 2001 * @qdisc_hash: qdisc hash table 2002 * @watchdog_timeo: Represents the timeout that is used by 2003 * the watchdog (see dev_watchdog()) 2004 * @watchdog_lock: protect watchdog_ref_held 2005 * @watchdog_ref_held: True if the watchdog device ref is taken. 2006 * @watchdog_timer: List of timers 2007 * 2008 * @proto_down_reason: reason a netdev interface is held down 2009 * @pcpu_refcnt: Number of references to this device 2010 * @dev_refcnt: Number of references to this device 2011 * @refcnt_tracker: Tracker directory for tracked references to this device 2012 * @todo_list: Delayed register/unregister 2013 * @link_watch_list: XXX: need comments on this one 2014 * 2015 * @reg_state: Register/unregister state machine 2016 * @dismantle: Device is going to be freed 2017 * @needs_free_netdev: Should unregister perform free_netdev? 2018 * @priv_destructor: Called from unregister 2019 * @npinfo: XXX: need comments on this one 2020 * @nd_net: Network namespace this network device is inside 2021 * protected by @lock 2022 * 2023 * @ml_priv: Mid-layer private 2024 * @ml_priv_type: Mid-layer private type 2025 * 2026 * @pcpu_stat_type: Type of device statistics which the core should 2027 * allocate/free: none, lstats, tstats, dstats. none 2028 * means the driver is handling statistics allocation/ 2029 * freeing internally. 2030 * @lstats: Loopback statistics: packets, bytes 2031 * @tstats: Tunnel statistics: RX/TX packets, RX/TX bytes 2032 * @dstats: Dummy statistics: RX/TX/drop packets, RX/TX bytes 2033 * 2034 * @garp_port: GARP 2035 * @mrp_port: MRP 2036 * 2037 * @dm_private: Drop monitor private 2038 * 2039 * @dev: Class/net/name entry 2040 * @sysfs_groups: Space for optional device, statistics and wireless 2041 * sysfs groups 2042 * 2043 * @sysfs_rx_queue_group: Space for optional per-rx queue attributes 2044 * @rtnl_link_ops: Rtnl_link_ops 2045 * @stat_ops: Optional ops for queue-aware statistics 2046 * @queue_mgmt_ops: Optional ops for queue management 2047 * 2048 * @gso_max_size: Maximum size of generic segmentation offload 2049 * @tso_max_size: Device (as in HW) limit on the max TSO request size 2050 * @gso_max_segs: Maximum number of segments that can be passed to the 2051 * NIC for GSO 2052 * @tso_max_segs: Device (as in HW) limit on the max TSO segment count 2053 * @gso_ipv4_max_size: Maximum size of generic segmentation offload, 2054 * for IPv4. 2055 * 2056 * @dcbnl_ops: Data Center Bridging netlink ops 2057 * @num_tc: Number of traffic classes in the net device 2058 * @tc_to_txq: XXX: need comments on this one 2059 * @prio_tc_map: XXX: need comments on this one 2060 * 2061 * @fcoe_ddp_xid: Max exchange id for FCoE LRO by ddp 2062 * 2063 * @priomap: XXX: need comments on this one 2064 * @link_topo: Physical link topology tracking attached PHYs 2065 * @phydev: Physical device may attach itself 2066 * for hardware timestamping 2067 * @sfp_bus: attached &struct sfp_bus structure. 2068 * 2069 * @qdisc_tx_busylock: lockdep class annotating Qdisc->busylock spinlock 2070 * 2071 * @proto_down: protocol port state information can be sent to the 2072 * switch driver and used to set the phys state of the 2073 * switch port. 2074 * 2075 * @irq_affinity_auto: driver wants the core to store and re-assign the IRQ 2076 * affinity. Set by netif_enable_irq_affinity(), then 2077 * the driver must create a persistent napi by 2078 * netif_napi_add_config() and finally bind the napi to 2079 * IRQ (via netif_napi_set_irq()). 2080 * 2081 * @rx_cpu_rmap_auto: driver wants the core to manage the ARFS rmap. 2082 * Set by calling netif_enable_cpu_rmap(). 2083 * 2084 * @see_all_hwtstamp_requests: device wants to see calls to 2085 * ndo_hwtstamp_set() for all timestamp requests 2086 * regardless of source, even if those aren't 2087 * HWTSTAMP_SOURCE_NETDEV 2088 * @change_proto_down: device supports setting carrier via IFLA_PROTO_DOWN 2089 * @netns_immutable: interface can't change network namespaces 2090 * @fcoe_mtu: device supports maximum FCoE MTU, 2158 bytes 2091 * 2092 * @net_notifier_list: List of per-net netdev notifier block 2093 * that follow this device when it is moved 2094 * to another network namespace. 2095 * 2096 * @macsec_ops: MACsec offloading ops 2097 * 2098 * @udp_tunnel_nic_info: static structure describing the UDP tunnel 2099 * offload capabilities of the device 2100 * @udp_tunnel_nic: UDP tunnel offload state 2101 * @ethtool: ethtool related state 2102 * @xdp_state: stores info on attached XDP BPF programs 2103 * 2104 * @nested_level: Used as a parameter of spin_lock_nested() of 2105 * dev->addr_list_lock. 2106 * @unlink_list: As netif_addr_lock() can be called recursively, 2107 * keep a list of interfaces to be deleted. 2108 * @gro_max_size: Maximum size of aggregated packet in generic 2109 * receive offload (GRO) 2110 * @gro_ipv4_max_size: Maximum size of aggregated packet in generic 2111 * receive offload (GRO), for IPv4. 2112 * @xdp_zc_max_segs: Maximum number of segments supported by AF_XDP 2113 * zero copy driver 2114 * 2115 * @dev_addr_shadow: Copy of @dev_addr to catch direct writes. 2116 * @linkwatch_dev_tracker: refcount tracker used by linkwatch. 2117 * @watchdog_dev_tracker: refcount tracker used by watchdog. 2118 * @dev_registered_tracker: tracker for reference held while 2119 * registered 2120 * @offload_xstats_l3: L3 HW stats for this netdevice. 2121 * 2122 * @devlink_port: Pointer to related devlink port structure. 2123 * Assigned by a driver before netdev registration using 2124 * SET_NETDEV_DEVLINK_PORT macro. This pointer is static 2125 * during the time netdevice is registered. 2126 * 2127 * @dpll_pin: Pointer to the SyncE source pin of a DPLL subsystem, 2128 * where the clock is recovered. 2129 * 2130 * @max_pacing_offload_horizon: max EDT offload horizon in nsec. 2131 * @napi_config: An array of napi_config structures containing per-NAPI 2132 * settings. 2133 * @num_napi_configs: number of allocated NAPI config structs, 2134 * always >= max(num_rx_queues, num_tx_queues). 2135 * @gro_flush_timeout: timeout for GRO layer in NAPI 2136 * @napi_defer_hard_irqs: If not zero, provides a counter that would 2137 * allow to avoid NIC hard IRQ, on busy queues. 2138 * 2139 * @neighbours: List heads pointing to this device's neighbours' 2140 * dev_list, one per address-family. 2141 * @hwprov: Tracks which PTP performs hardware packet time stamping. 2142 * 2143 * FIXME: cleanup struct net_device such that network protocol info 2144 * moves out. 2145 */ 2146 2147 struct net_device { 2148 /* Cacheline organization can be found documented in 2149 * Documentation/networking/net_cachelines/net_device.rst. 2150 * Please update the document when adding new fields. 2151 */ 2152 2153 /* TX read-mostly hotpath */ 2154 __cacheline_group_begin(net_device_read_tx); 2155 struct_group(priv_flags_fast, 2156 unsigned long priv_flags:32; 2157 unsigned long lltx:1; 2158 unsigned long netmem_tx:2; 2159 ); 2160 const struct net_device_ops *netdev_ops; 2161 const struct header_ops *header_ops; 2162 struct netdev_queue *_tx; 2163 netdev_features_t gso_partial_features; 2164 unsigned int real_num_tx_queues; 2165 unsigned int gso_max_size; 2166 unsigned int gso_ipv4_max_size; 2167 u16 gso_max_segs; 2168 s16 num_tc; 2169 /* Note : dev->mtu is often read without holding a lock. 2170 * Writers usually hold RTNL. 2171 * It is recommended to use READ_ONCE() to annotate the reads, 2172 * and to use WRITE_ONCE() to annotate the writes. 2173 */ 2174 unsigned int mtu; 2175 unsigned short needed_headroom; 2176 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE]; 2177 #ifdef CONFIG_XPS 2178 struct xps_dev_maps __rcu *xps_maps[XPS_MAPS_MAX]; 2179 #endif 2180 #ifdef CONFIG_NETFILTER_EGRESS 2181 struct nf_hook_entries __rcu *nf_hooks_egress; 2182 #endif 2183 #ifdef CONFIG_NET_XGRESS 2184 struct bpf_mprog_entry __rcu *tcx_egress; 2185 #endif 2186 __cacheline_group_end(net_device_read_tx); 2187 2188 /* TXRX read-mostly hotpath */ 2189 __cacheline_group_begin(net_device_read_txrx); 2190 union { 2191 struct pcpu_lstats __percpu *lstats; 2192 struct pcpu_sw_netstats __percpu *tstats; 2193 struct pcpu_dstats __percpu *dstats; 2194 }; 2195 unsigned long state; 2196 unsigned int flags; 2197 unsigned short hard_header_len; 2198 enum netdev_stat_type pcpu_stat_type:8; 2199 netdev_features_t features; 2200 struct inet6_dev __rcu *ip6_ptr; 2201 __cacheline_group_end(net_device_read_txrx); 2202 2203 /* RX read-mostly hotpath */ 2204 __cacheline_group_begin(net_device_read_rx); 2205 struct bpf_prog __rcu *xdp_prog; 2206 struct list_head ptype_specific; 2207 int ifindex; 2208 unsigned int real_num_rx_queues; 2209 struct netdev_rx_queue *_rx; 2210 unsigned int gro_max_size; 2211 unsigned int gro_ipv4_max_size; 2212 rx_handler_func_t __rcu *rx_handler; 2213 void __rcu *rx_handler_data; 2214 possible_net_t nd_net; 2215 #ifdef CONFIG_NETPOLL 2216 struct netpoll_info __rcu *npinfo; 2217 #endif 2218 #ifdef CONFIG_NET_XGRESS 2219 struct bpf_mprog_entry __rcu *tcx_ingress; 2220 #endif 2221 __cacheline_group_end(net_device_read_rx); 2222 2223 char name[IFNAMSIZ]; 2224 struct netdev_name_node *name_node; 2225 struct dev_ifalias __rcu *ifalias; 2226 /* 2227 * I/O specific fields 2228 * FIXME: Merge these and struct ifmap into one 2229 */ 2230 unsigned long mem_end; 2231 unsigned long mem_start; 2232 unsigned long base_addr; 2233 2234 /* 2235 * Some hardware also needs these fields (state,dev_list, 2236 * napi_list,unreg_list,close_list) but they are not 2237 * part of the usual set specified in Space.c. 2238 */ 2239 2240 2241 struct list_head dev_list; 2242 struct list_head napi_list; 2243 struct list_head unreg_list; 2244 struct list_head close_list; 2245 struct list_head ptype_all; 2246 2247 struct { 2248 struct list_head upper; 2249 struct list_head lower; 2250 } adj_list; 2251 2252 /* Read-mostly cache-line for fast-path access */ 2253 xdp_features_t xdp_features; 2254 const struct xdp_metadata_ops *xdp_metadata_ops; 2255 const struct xsk_tx_metadata_ops *xsk_tx_metadata_ops; 2256 unsigned short gflags; 2257 2258 unsigned short needed_tailroom; 2259 2260 netdev_features_t hw_features; 2261 netdev_features_t wanted_features; 2262 netdev_features_t vlan_features; 2263 netdev_features_t hw_enc_features; 2264 netdev_features_t mpls_features; 2265 netdev_features_t mangleid_features; 2266 2267 unsigned int min_mtu; 2268 unsigned int max_mtu; 2269 unsigned short type; 2270 unsigned char min_header_len; 2271 unsigned char name_assign_type; 2272 2273 int group; 2274 2275 struct net_device_stats stats; /* not used by modern drivers */ 2276 2277 struct net_device_core_stats __percpu *core_stats; 2278 2279 /* Stats to monitor link on/off, flapping */ 2280 atomic_t carrier_up_count; 2281 atomic_t carrier_down_count; 2282 2283 #ifdef CONFIG_WIRELESS_EXT 2284 const struct iw_handler_def *wireless_handlers; 2285 #endif 2286 const struct ethtool_ops *ethtool_ops; 2287 #ifdef CONFIG_NET_L3_MASTER_DEV 2288 const struct l3mdev_ops *l3mdev_ops; 2289 #endif 2290 #if IS_ENABLED(CONFIG_IPV6) 2291 const struct ndisc_ops *ndisc_ops; 2292 #endif 2293 2294 #ifdef CONFIG_XFRM_OFFLOAD 2295 const struct xfrmdev_ops *xfrmdev_ops; 2296 #endif 2297 2298 #if IS_ENABLED(CONFIG_TLS_DEVICE) 2299 const struct tlsdev_ops *tlsdev_ops; 2300 #endif 2301 2302 unsigned int operstate; 2303 unsigned char link_mode; 2304 2305 unsigned char if_port; 2306 unsigned char dma; 2307 2308 /* Interface address info. */ 2309 unsigned char perm_addr[MAX_ADDR_LEN]; 2310 unsigned char addr_assign_type; 2311 unsigned char addr_len; 2312 unsigned char upper_level; 2313 unsigned char lower_level; 2314 u8 threaded; 2315 2316 unsigned short neigh_priv_len; 2317 unsigned short dev_id; 2318 unsigned short dev_port; 2319 int irq; 2320 u32 priv_len; 2321 2322 spinlock_t addr_list_lock; 2323 2324 struct netdev_hw_addr_list uc; 2325 struct netdev_hw_addr_list mc; 2326 struct netdev_hw_addr_list dev_addrs; 2327 2328 #ifdef CONFIG_SYSFS 2329 struct kset *queues_kset; 2330 #endif 2331 #ifdef CONFIG_LOCKDEP 2332 struct list_head unlink_list; 2333 #endif 2334 unsigned int promiscuity; 2335 unsigned int allmulti; 2336 bool uc_promisc; 2337 struct list_head work_node; 2338 netdevice_tracker work_tracker; 2339 unsigned long work_pending; 2340 unsigned long work_core_pending; 2341 struct netdev_hw_addr_list rx_mode_addr_cache; 2342 struct timer_list rx_mode_retry_timer; 2343 unsigned int rx_mode_retry_count; 2344 #ifdef CONFIG_LOCKDEP 2345 unsigned char nested_level; 2346 #endif 2347 2348 2349 /* Protocol-specific pointers */ 2350 struct in_device __rcu *ip_ptr; 2351 /** @fib_nh_head: nexthops associated with this netdev */ 2352 struct hlist_head fib_nh_head; 2353 2354 #if IS_ENABLED(CONFIG_VLAN_8021Q) 2355 struct vlan_info __rcu *vlan_info; 2356 #endif 2357 #if IS_ENABLED(CONFIG_NET_DSA) 2358 struct dsa_port *dsa_ptr; 2359 #endif 2360 #if IS_ENABLED(CONFIG_TIPC) 2361 struct tipc_bearer __rcu *tipc_ptr; 2362 #endif 2363 #if IS_ENABLED(CONFIG_CFG80211) 2364 struct wireless_dev *ieee80211_ptr; 2365 #endif 2366 #if IS_ENABLED(CONFIG_IEEE802154) || IS_ENABLED(CONFIG_6LOWPAN) 2367 struct wpan_dev *ieee802154_ptr; 2368 #endif 2369 #if IS_ENABLED(CONFIG_MPLS_ROUTING) 2370 struct mpls_dev __rcu *mpls_ptr; 2371 #endif 2372 #if IS_ENABLED(CONFIG_MCTP) 2373 struct mctp_dev __rcu *mctp_ptr; 2374 #endif 2375 #if IS_ENABLED(CONFIG_INET_PSP) 2376 struct psp_dev __rcu *psp_dev; 2377 #endif 2378 2379 /* 2380 * Cache lines mostly used on receive path (including eth_type_trans()) 2381 */ 2382 /* Interface address info used in eth_type_trans() */ 2383 const unsigned char *dev_addr; 2384 2385 unsigned int num_rx_queues; 2386 #define GRO_LEGACY_MAX_SIZE 65536u 2387 /* TCP minimal MSS is 8 (TCP_MIN_GSO_SIZE), 2388 * and shinfo->gso_segs is a 16bit field. 2389 */ 2390 #define GRO_MAX_SIZE (8 * 65535u) 2391 unsigned int xdp_zc_max_segs; 2392 struct netdev_queue __rcu *ingress_queue; 2393 #ifdef CONFIG_NETFILTER_INGRESS 2394 struct nf_hook_entries __rcu *nf_hooks_ingress; 2395 #endif 2396 2397 unsigned char broadcast[MAX_ADDR_LEN]; 2398 #ifdef CONFIG_RFS_ACCEL 2399 struct cpu_rmap *rx_cpu_rmap; 2400 #endif 2401 struct hlist_node index_hlist; 2402 2403 /* 2404 * Cache lines mostly used on transmit path 2405 */ 2406 unsigned int num_tx_queues; 2407 struct Qdisc __rcu *qdisc; 2408 unsigned int tx_queue_len; 2409 spinlock_t tx_global_lock; 2410 2411 struct xdp_dev_bulk_queue __percpu *xdp_bulkq; 2412 2413 #ifdef CONFIG_NET_SCHED 2414 DECLARE_HASHTABLE (qdisc_hash, 4); 2415 #endif 2416 /* These may be needed for future network-power-down code. */ 2417 struct timer_list watchdog_timer; 2418 int watchdog_timeo; 2419 spinlock_t watchdog_lock; 2420 bool watchdog_ref_held; 2421 2422 u32 proto_down_reason; 2423 2424 struct list_head todo_list; 2425 2426 #ifdef CONFIG_PCPU_DEV_REFCNT 2427 int __percpu *pcpu_refcnt; 2428 #else 2429 refcount_t dev_refcnt; 2430 #endif 2431 struct ref_tracker_dir refcnt_tracker; 2432 2433 struct list_head link_watch_list; 2434 2435 u8 reg_state; 2436 2437 bool dismantle; 2438 2439 /** @moving_ns: device is changing netns, protected by @lock */ 2440 bool moving_ns; 2441 /** @rtnl_link_initializing: Device being created, suppress events */ 2442 bool rtnl_link_initializing; 2443 2444 bool needs_free_netdev; 2445 void (*priv_destructor)(struct net_device *dev); 2446 2447 /* mid-layer private */ 2448 void *ml_priv; 2449 enum netdev_ml_priv_type ml_priv_type; 2450 2451 #if IS_ENABLED(CONFIG_GARP) 2452 struct garp_port __rcu *garp_port; 2453 #endif 2454 #if IS_ENABLED(CONFIG_MRP) 2455 struct mrp_port __rcu *mrp_port; 2456 #endif 2457 #if IS_ENABLED(CONFIG_NET_DROP_MONITOR) 2458 struct dm_hw_stat_delta __rcu *dm_private; 2459 #endif 2460 struct device dev; 2461 const struct attribute_group *sysfs_groups[5]; 2462 const struct attribute_group *sysfs_rx_queue_group; 2463 2464 const struct rtnl_link_ops *rtnl_link_ops; 2465 2466 const struct netdev_stat_ops *stat_ops; 2467 2468 const struct netdev_queue_mgmt_ops *queue_mgmt_ops; 2469 2470 /* for setting kernel sock attribute on TCP connection setup */ 2471 #define GSO_MAX_SEGS 65535u 2472 #define GSO_LEGACY_MAX_SIZE 65536u 2473 /* TCP minimal MSS is 8 (TCP_MIN_GSO_SIZE), 2474 * and shinfo->gso_segs is a 16bit field. 2475 */ 2476 #define GSO_MAX_SIZE (8 * GSO_MAX_SEGS) 2477 2478 #define TSO_LEGACY_MAX_SIZE 65536 2479 #define TSO_MAX_SIZE UINT_MAX 2480 unsigned int tso_max_size; 2481 #define TSO_MAX_SEGS U16_MAX 2482 u16 tso_max_segs; 2483 2484 #ifdef CONFIG_DCB 2485 const struct dcbnl_rtnl_ops *dcbnl_ops; 2486 #endif 2487 u8 prio_tc_map[TC_BITMASK + 1]; 2488 2489 #if IS_ENABLED(CONFIG_FCOE) 2490 unsigned int fcoe_ddp_xid; 2491 #endif 2492 #if IS_ENABLED(CONFIG_CGROUP_NET_PRIO) 2493 struct netprio_map __rcu *priomap; 2494 #endif 2495 struct phy_link_topology *link_topo; 2496 struct phy_device *phydev; 2497 struct sfp_bus *sfp_bus; 2498 struct lock_class_key *qdisc_tx_busylock; 2499 bool proto_down; 2500 bool irq_affinity_auto; 2501 bool rx_cpu_rmap_auto; 2502 2503 /* priv_flags_slow, ungrouped to save space */ 2504 unsigned long see_all_hwtstamp_requests:1; 2505 unsigned long change_proto_down:1; 2506 unsigned long netns_immutable:1; 2507 unsigned long fcoe_mtu:1; 2508 2509 struct list_head net_notifier_list; 2510 2511 #if IS_ENABLED(CONFIG_MACSEC) 2512 /* MACsec management functions */ 2513 const struct macsec_ops *macsec_ops; 2514 #endif 2515 const struct udp_tunnel_nic_info *udp_tunnel_nic_info; 2516 struct udp_tunnel_nic *udp_tunnel_nic; 2517 2518 /** @cfg: net_device queue-related configuration */ 2519 struct netdev_config *cfg; 2520 /** 2521 * @cfg_pending: same as @cfg but when device is being actively 2522 * reconfigured includes any changes to the configuration 2523 * requested by the user, but which may or may not be rejected. 2524 */ 2525 struct netdev_config *cfg_pending; 2526 struct ethtool_netdev_state *ethtool; 2527 2528 /* protected by rtnl_lock */ 2529 struct bpf_xdp_entity xdp_state[__MAX_XDP_MODE]; 2530 2531 u8 dev_addr_shadow[MAX_ADDR_LEN]; 2532 netdevice_tracker linkwatch_dev_tracker; 2533 netdevice_tracker watchdog_dev_tracker; 2534 netdevice_tracker dev_registered_tracker; 2535 struct rtnl_hw_stats64 *offload_xstats_l3; 2536 2537 struct devlink_port *devlink_port; 2538 2539 #if IS_ENABLED(CONFIG_DPLL) 2540 struct dpll_pin __rcu *dpll_pin; 2541 #endif 2542 #if IS_ENABLED(CONFIG_PAGE_POOL) 2543 /** @page_pools: page pools created for this netdevice */ 2544 struct hlist_head page_pools; 2545 #endif 2546 2547 /** @irq_moder: dim parameters used if IS_ENABLED(CONFIG_DIMLIB). */ 2548 struct dim_irq_moder *irq_moder; 2549 2550 u64 max_pacing_offload_horizon; 2551 struct napi_config *napi_config; 2552 u32 num_napi_configs; 2553 u32 napi_defer_hard_irqs; 2554 unsigned long gro_flush_timeout; 2555 2556 /** 2557 * @up: copy of @state's IFF_UP, but safe to read with just @lock. 2558 * May report false negatives while the device is being opened 2559 * or closed (@lock does not protect .ndo_open, or .ndo_close). 2560 */ 2561 bool up; 2562 2563 /** 2564 * @request_ops_lock: request the core to run all @netdev_ops and 2565 * @ethtool_ops under the @lock. 2566 */ 2567 bool request_ops_lock; 2568 2569 /** 2570 * @lock: netdev-scope lock, protects a small selection of fields. 2571 * Should always be taken using netdev_lock() / netdev_unlock() helpers. 2572 * Drivers are free to use it for other protection. 2573 * 2574 * For the drivers that implement shaper or queue API, the scope 2575 * of this lock is expanded to cover most ndo/queue/ethtool/sysfs 2576 * operations. Drivers may opt-in to this behavior by setting 2577 * @request_ops_lock. 2578 * 2579 * @lock protection mixes with rtnl_lock in multiple ways, fields are 2580 * either: 2581 * 2582 * - simply protected by the instance @lock; 2583 * 2584 * - double protected - writers hold both locks, readers hold either; 2585 * 2586 * - ops protected - protected by the lock held around the NDOs 2587 * and other callbacks, that is the instance lock on devices for 2588 * which netdev_need_ops_lock() returns true, otherwise by rtnl_lock; 2589 * 2590 * - double ops protected - always protected by rtnl_lock but for 2591 * devices for which netdev_need_ops_lock() returns true - also 2592 * the instance lock. 2593 * 2594 * Simply protects: 2595 * @gro_flush_timeout, @napi_defer_hard_irqs, @napi_list, 2596 * @net_shaper_hierarchy, @reg_state, @threaded 2597 * 2598 * Double protects: 2599 * @up, @moving_ns, @nd_net, @xdp_features 2600 * 2601 * Ops protects: 2602 * @cfg, @cfg_pending, @ethtool, @hwprov 2603 * 2604 * Double ops protects: 2605 * @real_num_rx_queues, @real_num_tx_queues 2606 * 2607 * Also protects some fields in: 2608 * struct napi_struct, struct netdev_queue, struct netdev_rx_queue 2609 * 2610 * Ordering: 2611 * 2612 * - take after rtnl_lock 2613 * 2614 * - for the case of netdev queue leasing, the netdev-scope lock is 2615 * taken for both the virtual and the physical device; to prevent 2616 * deadlocks, the virtual device's lock must always be acquired 2617 * before the physical device's (see netdev_nl_queue_create_doit) 2618 */ 2619 struct mutex lock; 2620 2621 #if IS_ENABLED(CONFIG_NET_SHAPER) 2622 /** 2623 * @net_shaper_hierarchy: data tracking the current shaper status 2624 * see include/net/net_shapers.h 2625 */ 2626 struct net_shaper_hierarchy *net_shaper_hierarchy; 2627 #endif 2628 2629 struct hlist_head neighbours[NEIGH_NR_TABLES]; 2630 2631 struct hwtstamp_provider __rcu *hwprov; 2632 2633 u8 priv[] ____cacheline_aligned 2634 __counted_by(priv_len); 2635 } ____cacheline_aligned; 2636 #define to_net_dev(d) container_of(d, struct net_device, dev) 2637 2638 /* 2639 * Driver should use this to assign devlink port instance to a netdevice 2640 * before it registers the netdevice. Therefore devlink_port is static 2641 * during the netdev lifetime after it is registered. 2642 */ 2643 #define SET_NETDEV_DEVLINK_PORT(dev, port) \ 2644 ({ \ 2645 WARN_ON((dev)->reg_state != NETREG_UNINITIALIZED); \ 2646 ((dev)->devlink_port = (port)); \ 2647 }) 2648 2649 static inline bool netif_elide_gro(const struct net_device *dev) 2650 { 2651 if (!(dev->features & NETIF_F_GRO) || dev->xdp_prog) 2652 return true; 2653 return false; 2654 } 2655 2656 #define NETDEV_ALIGN 32 2657 2658 static inline 2659 int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio) 2660 { 2661 return dev->prio_tc_map[prio & TC_BITMASK]; 2662 } 2663 2664 static inline 2665 int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc) 2666 { 2667 if (tc >= dev->num_tc) 2668 return -EINVAL; 2669 2670 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK; 2671 return 0; 2672 } 2673 2674 int netdev_txq_to_tc(struct net_device *dev, unsigned int txq); 2675 void netdev_reset_tc(struct net_device *dev); 2676 int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset); 2677 int netdev_set_num_tc(struct net_device *dev, u8 num_tc); 2678 2679 static inline 2680 int netdev_get_num_tc(struct net_device *dev) 2681 { 2682 return dev->num_tc; 2683 } 2684 2685 static inline void net_prefetch(void *p) 2686 { 2687 prefetch(p); 2688 #if L1_CACHE_BYTES < 128 2689 prefetch((u8 *)p + L1_CACHE_BYTES); 2690 #endif 2691 } 2692 2693 static inline void net_prefetchw(void *p) 2694 { 2695 prefetchw(p); 2696 #if L1_CACHE_BYTES < 128 2697 prefetchw((u8 *)p + L1_CACHE_BYTES); 2698 #endif 2699 } 2700 2701 void netdev_unbind_sb_channel(struct net_device *dev, 2702 struct net_device *sb_dev); 2703 int netdev_bind_sb_channel_queue(struct net_device *dev, 2704 struct net_device *sb_dev, 2705 u8 tc, u16 count, u16 offset); 2706 int netdev_set_sb_channel(struct net_device *dev, u16 channel); 2707 static inline int netdev_get_sb_channel(struct net_device *dev) 2708 { 2709 return max_t(int, -dev->num_tc, 0); 2710 } 2711 2712 static inline 2713 struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev, 2714 unsigned int index) 2715 { 2716 DEBUG_NET_WARN_ON_ONCE(index >= dev->num_tx_queues); 2717 return &dev->_tx[index]; 2718 } 2719 2720 static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev, 2721 const struct sk_buff *skb) 2722 { 2723 return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb)); 2724 } 2725 2726 static inline void netdev_for_each_tx_queue(struct net_device *dev, 2727 void (*f)(struct net_device *, 2728 struct netdev_queue *, 2729 void *), 2730 void *arg) 2731 { 2732 unsigned int i; 2733 2734 for (i = 0; i < dev->num_tx_queues; i++) 2735 f(dev, &dev->_tx[i], arg); 2736 } 2737 2738 u16 netdev_pick_tx(struct net_device *dev, struct sk_buff *skb, 2739 struct net_device *sb_dev); 2740 struct netdev_queue *netdev_core_pick_tx(struct net_device *dev, 2741 struct sk_buff *skb, 2742 struct net_device *sb_dev); 2743 2744 /* returns the headroom that the master device needs to take in account 2745 * when forwarding to this dev 2746 */ 2747 static inline unsigned netdev_get_fwd_headroom(struct net_device *dev) 2748 { 2749 return dev->priv_flags & IFF_PHONY_HEADROOM ? 0 : dev->needed_headroom; 2750 } 2751 2752 static inline void netdev_set_rx_headroom(struct net_device *dev, int new_hr) 2753 { 2754 if (dev->netdev_ops->ndo_set_rx_headroom) 2755 dev->netdev_ops->ndo_set_rx_headroom(dev, new_hr); 2756 } 2757 2758 /* set the device rx headroom to the dev's default */ 2759 static inline void netdev_reset_rx_headroom(struct net_device *dev) 2760 { 2761 netdev_set_rx_headroom(dev, -1); 2762 } 2763 2764 static inline void *netdev_get_ml_priv(struct net_device *dev, 2765 enum netdev_ml_priv_type type) 2766 { 2767 if (dev->ml_priv_type != type) 2768 return NULL; 2769 2770 return dev->ml_priv; 2771 } 2772 2773 static inline void netdev_set_ml_priv(struct net_device *dev, 2774 void *ml_priv, 2775 enum netdev_ml_priv_type type) 2776 { 2777 WARN(dev->ml_priv_type && dev->ml_priv_type != type, 2778 "Overwriting already set ml_priv_type (%u) with different ml_priv_type (%u)!\n", 2779 dev->ml_priv_type, type); 2780 WARN(!dev->ml_priv_type && dev->ml_priv, 2781 "Overwriting already set ml_priv and ml_priv_type is ML_PRIV_NONE!\n"); 2782 2783 dev->ml_priv = ml_priv; 2784 dev->ml_priv_type = type; 2785 } 2786 2787 /* 2788 * Net namespace inlines 2789 */ 2790 static inline 2791 struct net *dev_net(const struct net_device *dev) 2792 { 2793 return read_pnet(&dev->nd_net); 2794 } 2795 2796 static inline 2797 struct net *dev_net_rcu(const struct net_device *dev) 2798 { 2799 return read_pnet_rcu(&dev->nd_net); 2800 } 2801 2802 static inline 2803 void dev_net_set(struct net_device *dev, struct net *net) 2804 { 2805 write_pnet(&dev->nd_net, net); 2806 } 2807 2808 /** 2809 * netdev_priv - access network device private data 2810 * @dev: network device 2811 * 2812 * Get network device private data 2813 */ 2814 static inline void *netdev_priv(const struct net_device *dev) 2815 { 2816 return (void *)dev->priv; 2817 } 2818 2819 /** 2820 * netdev_from_priv() - get network device from priv 2821 * @priv: network device private data 2822 * 2823 * Returns: net_device to which @priv belongs 2824 */ 2825 static inline struct net_device *netdev_from_priv(const void *priv) 2826 { 2827 return container_of(priv, struct net_device, priv); 2828 } 2829 2830 /* Set the sysfs physical device reference for the network logical device 2831 * if set prior to registration will cause a symlink during initialization. 2832 */ 2833 #define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev)) 2834 2835 /* Set the sysfs device type for the network logical device to allow 2836 * fine-grained identification of different network device types. For 2837 * example Ethernet, Wireless LAN, Bluetooth, WiMAX etc. 2838 */ 2839 #define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype)) 2840 2841 void netif_queue_set_napi(struct net_device *dev, unsigned int queue_index, 2842 enum netdev_queue_type type, 2843 struct napi_struct *napi); 2844 2845 static inline void netdev_lock(struct net_device *dev) 2846 { 2847 mutex_lock(&dev->lock); 2848 } 2849 2850 static inline void netdev_unlock(struct net_device *dev) 2851 { 2852 mutex_unlock(&dev->lock); 2853 } 2854 /* Additional netdev_lock()-related helpers are in net/netdev_lock.h */ 2855 2856 void netif_napi_set_irq_locked(struct napi_struct *napi, int irq); 2857 2858 static inline void netif_napi_set_irq(struct napi_struct *napi, int irq) 2859 { 2860 netdev_lock(napi->dev); 2861 netif_napi_set_irq_locked(napi, irq); 2862 netdev_unlock(napi->dev); 2863 } 2864 2865 /* Default NAPI poll() weight 2866 * Device drivers are strongly advised to not use bigger value 2867 */ 2868 #define NAPI_POLL_WEIGHT 64 2869 2870 void netif_napi_add_weight_locked(struct net_device *dev, 2871 struct napi_struct *napi, 2872 int (*poll)(struct napi_struct *, int), 2873 int weight); 2874 2875 static inline void 2876 netif_napi_add_weight(struct net_device *dev, struct napi_struct *napi, 2877 int (*poll)(struct napi_struct *, int), int weight) 2878 { 2879 netdev_lock(dev); 2880 netif_napi_add_weight_locked(dev, napi, poll, weight); 2881 netdev_unlock(dev); 2882 } 2883 2884 /** 2885 * netif_napi_add() - initialize a NAPI context 2886 * @dev: network device 2887 * @napi: NAPI context 2888 * @poll: polling function 2889 * 2890 * netif_napi_add() must be used to initialize a NAPI context prior to calling 2891 * *any* of the other NAPI-related functions. 2892 */ 2893 static inline void 2894 netif_napi_add(struct net_device *dev, struct napi_struct *napi, 2895 int (*poll)(struct napi_struct *, int)) 2896 { 2897 netif_napi_add_weight(dev, napi, poll, NAPI_POLL_WEIGHT); 2898 } 2899 2900 static inline void 2901 netif_napi_add_locked(struct net_device *dev, struct napi_struct *napi, 2902 int (*poll)(struct napi_struct *, int)) 2903 { 2904 netif_napi_add_weight_locked(dev, napi, poll, NAPI_POLL_WEIGHT); 2905 } 2906 2907 static inline void 2908 netif_napi_add_tx_weight(struct net_device *dev, 2909 struct napi_struct *napi, 2910 int (*poll)(struct napi_struct *, int), 2911 int weight) 2912 { 2913 set_bit(NAPI_STATE_NO_BUSY_POLL, &napi->state); 2914 netif_napi_add_weight(dev, napi, poll, weight); 2915 } 2916 2917 static inline void 2918 netif_napi_add_config_locked(struct net_device *dev, struct napi_struct *napi, 2919 int (*poll)(struct napi_struct *, int), int index) 2920 { 2921 napi->index = index; 2922 napi->config = &dev->napi_config[index]; 2923 netif_napi_add_weight_locked(dev, napi, poll, NAPI_POLL_WEIGHT); 2924 } 2925 2926 /** 2927 * netif_napi_add_config - initialize a NAPI context with persistent config 2928 * @dev: network device 2929 * @napi: NAPI context 2930 * @poll: polling function 2931 * @index: the NAPI index 2932 */ 2933 static inline void 2934 netif_napi_add_config(struct net_device *dev, struct napi_struct *napi, 2935 int (*poll)(struct napi_struct *, int), int index) 2936 { 2937 netdev_lock(dev); 2938 netif_napi_add_config_locked(dev, napi, poll, index); 2939 netdev_unlock(dev); 2940 } 2941 2942 /** 2943 * netif_napi_add_tx() - initialize a NAPI context to be used for Tx only 2944 * @dev: network device 2945 * @napi: NAPI context 2946 * @poll: polling function 2947 * 2948 * This variant of netif_napi_add() should be used from drivers using NAPI 2949 * to exclusively poll a TX queue. 2950 * This will avoid we add it into napi_hash[], thus polluting this hash table. 2951 */ 2952 static inline void netif_napi_add_tx(struct net_device *dev, 2953 struct napi_struct *napi, 2954 int (*poll)(struct napi_struct *, int)) 2955 { 2956 netif_napi_add_tx_weight(dev, napi, poll, NAPI_POLL_WEIGHT); 2957 } 2958 2959 void __netif_napi_del_locked(struct napi_struct *napi); 2960 2961 /** 2962 * __netif_napi_del - remove a NAPI context 2963 * @napi: NAPI context 2964 * 2965 * Warning: caller must observe RCU grace period before freeing memory 2966 * containing @napi. Drivers might want to call this helper to combine 2967 * all the needed RCU grace periods into a single one. 2968 */ 2969 static inline void __netif_napi_del(struct napi_struct *napi) 2970 { 2971 netdev_lock(napi->dev); 2972 __netif_napi_del_locked(napi); 2973 netdev_unlock(napi->dev); 2974 } 2975 2976 static inline void netif_napi_del_locked(struct napi_struct *napi) 2977 { 2978 __netif_napi_del_locked(napi); 2979 synchronize_net(); 2980 } 2981 2982 /** 2983 * netif_napi_del - remove a NAPI context 2984 * @napi: NAPI context 2985 * 2986 * netif_napi_del() removes a NAPI context from the network device NAPI list 2987 */ 2988 static inline void netif_napi_del(struct napi_struct *napi) 2989 { 2990 __netif_napi_del(napi); 2991 synchronize_net(); 2992 } 2993 2994 int netif_enable_cpu_rmap(struct net_device *dev, unsigned int num_irqs); 2995 void netif_set_affinity_auto(struct net_device *dev); 2996 2997 struct packet_type { 2998 __be16 type; /* This is really htons(ether_type). */ 2999 bool ignore_outgoing; 3000 struct net_device *dev; /* NULL is wildcarded here */ 3001 netdevice_tracker dev_tracker; 3002 int (*func) (struct sk_buff *, 3003 struct net_device *, 3004 struct packet_type *, 3005 struct net_device *); 3006 void (*list_func) (struct list_head *, 3007 struct packet_type *, 3008 struct net_device *); 3009 bool (*id_match)(struct packet_type *ptype, 3010 struct sock *sk); 3011 struct net *af_packet_net; 3012 void *af_packet_priv; 3013 struct list_head list; 3014 }; 3015 3016 struct offload_callbacks { 3017 struct sk_buff *(*gso_segment)(struct sk_buff *skb, 3018 netdev_features_t features); 3019 struct sk_buff *(*gro_receive)(struct list_head *head, 3020 struct sk_buff *skb); 3021 int (*gro_complete)(struct sk_buff *skb, int nhoff); 3022 }; 3023 3024 struct packet_offload { 3025 __be16 type; /* This is really htons(ether_type). */ 3026 u16 priority; 3027 struct offload_callbacks callbacks; 3028 struct list_head list; 3029 }; 3030 3031 /* often modified stats are per-CPU, other are shared (netdev->stats) */ 3032 struct pcpu_sw_netstats { 3033 u64_stats_t rx_packets; 3034 u64_stats_t rx_bytes; 3035 u64_stats_t tx_packets; 3036 u64_stats_t tx_bytes; 3037 struct u64_stats_sync syncp; 3038 } __aligned(4 * sizeof(u64)); 3039 3040 struct pcpu_dstats { 3041 u64_stats_t rx_packets; 3042 u64_stats_t rx_bytes; 3043 u64_stats_t tx_packets; 3044 u64_stats_t tx_bytes; 3045 u64_stats_t rx_drops; 3046 u64_stats_t tx_drops; 3047 struct u64_stats_sync syncp; 3048 } __aligned(8 * sizeof(u64)); 3049 3050 struct pcpu_lstats { 3051 u64_stats_t packets; 3052 u64_stats_t bytes; 3053 struct u64_stats_sync syncp; 3054 } __aligned(2 * sizeof(u64)); 3055 3056 void dev_lstats_read(struct net_device *dev, u64 *packets, u64 *bytes); 3057 3058 static inline void dev_sw_netstats_rx_add(struct net_device *dev, unsigned int len) 3059 { 3060 struct pcpu_sw_netstats *tstats = this_cpu_ptr(dev->tstats); 3061 3062 u64_stats_update_begin(&tstats->syncp); 3063 u64_stats_add(&tstats->rx_bytes, len); 3064 u64_stats_inc(&tstats->rx_packets); 3065 u64_stats_update_end(&tstats->syncp); 3066 } 3067 3068 static inline void dev_sw_netstats_tx_add(struct net_device *dev, 3069 unsigned int packets, 3070 unsigned int len) 3071 { 3072 struct pcpu_sw_netstats *tstats = this_cpu_ptr(dev->tstats); 3073 3074 u64_stats_update_begin(&tstats->syncp); 3075 u64_stats_add(&tstats->tx_bytes, len); 3076 u64_stats_add(&tstats->tx_packets, packets); 3077 u64_stats_update_end(&tstats->syncp); 3078 } 3079 3080 static inline void dev_lstats_add(struct net_device *dev, unsigned int len) 3081 { 3082 struct pcpu_lstats *lstats = this_cpu_ptr(dev->lstats); 3083 3084 u64_stats_update_begin(&lstats->syncp); 3085 u64_stats_add(&lstats->bytes, len); 3086 u64_stats_inc(&lstats->packets); 3087 u64_stats_update_end(&lstats->syncp); 3088 } 3089 3090 static inline void dev_dstats_rx_add(struct net_device *dev, 3091 unsigned int len) 3092 { 3093 struct pcpu_dstats *dstats = this_cpu_ptr(dev->dstats); 3094 3095 u64_stats_update_begin(&dstats->syncp); 3096 u64_stats_inc(&dstats->rx_packets); 3097 u64_stats_add(&dstats->rx_bytes, len); 3098 u64_stats_update_end(&dstats->syncp); 3099 } 3100 3101 static inline void dev_dstats_rx_dropped(struct net_device *dev) 3102 { 3103 struct pcpu_dstats *dstats = this_cpu_ptr(dev->dstats); 3104 3105 u64_stats_update_begin(&dstats->syncp); 3106 u64_stats_inc(&dstats->rx_drops); 3107 u64_stats_update_end(&dstats->syncp); 3108 } 3109 3110 static inline void dev_dstats_rx_dropped_add(struct net_device *dev, 3111 unsigned int packets) 3112 { 3113 struct pcpu_dstats *dstats = this_cpu_ptr(dev->dstats); 3114 3115 u64_stats_update_begin(&dstats->syncp); 3116 u64_stats_add(&dstats->rx_drops, packets); 3117 u64_stats_update_end(&dstats->syncp); 3118 } 3119 3120 static inline void dev_dstats_tx_add(struct net_device *dev, 3121 unsigned int len) 3122 { 3123 struct pcpu_dstats *dstats = this_cpu_ptr(dev->dstats); 3124 3125 u64_stats_update_begin(&dstats->syncp); 3126 u64_stats_inc(&dstats->tx_packets); 3127 u64_stats_add(&dstats->tx_bytes, len); 3128 u64_stats_update_end(&dstats->syncp); 3129 } 3130 3131 static inline void dev_dstats_tx_dropped(struct net_device *dev) 3132 { 3133 struct pcpu_dstats *dstats = this_cpu_ptr(dev->dstats); 3134 3135 u64_stats_update_begin(&dstats->syncp); 3136 u64_stats_inc(&dstats->tx_drops); 3137 u64_stats_update_end(&dstats->syncp); 3138 } 3139 3140 #define __netdev_alloc_pcpu_stats(type, gfp) \ 3141 ({ \ 3142 typeof(type) __percpu *pcpu_stats = alloc_percpu_gfp(type, gfp);\ 3143 if (pcpu_stats) { \ 3144 int __cpu; \ 3145 for_each_possible_cpu(__cpu) { \ 3146 typeof(type) *stat; \ 3147 stat = per_cpu_ptr(pcpu_stats, __cpu); \ 3148 u64_stats_init(&stat->syncp); \ 3149 } \ 3150 } \ 3151 pcpu_stats; \ 3152 }) 3153 3154 #define netdev_alloc_pcpu_stats(type) \ 3155 __netdev_alloc_pcpu_stats(type, GFP_KERNEL) 3156 3157 #define devm_netdev_alloc_pcpu_stats(dev, type) \ 3158 ({ \ 3159 typeof(type) __percpu *pcpu_stats = devm_alloc_percpu(dev, type);\ 3160 if (pcpu_stats) { \ 3161 int __cpu; \ 3162 for_each_possible_cpu(__cpu) { \ 3163 typeof(type) *stat; \ 3164 stat = per_cpu_ptr(pcpu_stats, __cpu); \ 3165 u64_stats_init(&stat->syncp); \ 3166 } \ 3167 } \ 3168 pcpu_stats; \ 3169 }) 3170 3171 enum netdev_lag_tx_type { 3172 NETDEV_LAG_TX_TYPE_UNKNOWN, 3173 NETDEV_LAG_TX_TYPE_RANDOM, 3174 NETDEV_LAG_TX_TYPE_BROADCAST, 3175 NETDEV_LAG_TX_TYPE_ROUNDROBIN, 3176 NETDEV_LAG_TX_TYPE_ACTIVEBACKUP, 3177 NETDEV_LAG_TX_TYPE_HASH, 3178 }; 3179 3180 enum netdev_lag_hash { 3181 NETDEV_LAG_HASH_NONE, 3182 NETDEV_LAG_HASH_L2, 3183 NETDEV_LAG_HASH_L34, 3184 NETDEV_LAG_HASH_L23, 3185 NETDEV_LAG_HASH_E23, 3186 NETDEV_LAG_HASH_E34, 3187 NETDEV_LAG_HASH_VLAN_SRCMAC, 3188 NETDEV_LAG_HASH_UNKNOWN, 3189 }; 3190 3191 struct netdev_lag_upper_info { 3192 enum netdev_lag_tx_type tx_type; 3193 enum netdev_lag_hash hash_type; 3194 }; 3195 3196 struct netdev_lag_lower_state_info { 3197 u8 link_up : 1, 3198 tx_enabled : 1; 3199 }; 3200 3201 #include <linux/notifier.h> 3202 3203 /* netdevice notifier chain. Please remember to update netdev_cmd_to_name() 3204 * and the rtnetlink notification exclusion list in rtnetlink_event() when 3205 * adding new types. 3206 */ 3207 enum netdev_cmd { 3208 NETDEV_UP = 1, /* For now you can't veto a device up/down */ 3209 NETDEV_DOWN, 3210 NETDEV_REBOOT, /* Tell a protocol stack a network interface 3211 detected a hardware crash and restarted 3212 - we can use this eg to kick tcp sessions 3213 once done */ 3214 NETDEV_CHANGE, /* Notify device state change */ 3215 NETDEV_REGISTER, 3216 NETDEV_UNREGISTER, 3217 NETDEV_CHANGEMTU, /* notify after mtu change happened */ 3218 NETDEV_CHANGEADDR, /* notify after the address change */ 3219 NETDEV_PRE_CHANGEADDR, /* notify before the address change */ 3220 NETDEV_GOING_DOWN, 3221 NETDEV_CHANGENAME, 3222 NETDEV_FEAT_CHANGE, 3223 NETDEV_BONDING_FAILOVER, 3224 NETDEV_PRE_UP, 3225 NETDEV_PRE_TYPE_CHANGE, 3226 NETDEV_POST_TYPE_CHANGE, 3227 NETDEV_POST_INIT, 3228 NETDEV_PRE_UNINIT, 3229 NETDEV_RELEASE, 3230 NETDEV_NOTIFY_PEERS, 3231 NETDEV_JOIN, 3232 NETDEV_CHANGEUPPER, 3233 NETDEV_RESEND_IGMP, 3234 NETDEV_PRECHANGEMTU, /* notify before mtu change happened */ 3235 NETDEV_CHANGEINFODATA, 3236 NETDEV_BONDING_INFO, 3237 NETDEV_PRECHANGEUPPER, 3238 NETDEV_CHANGELOWERSTATE, 3239 NETDEV_UDP_TUNNEL_PUSH_INFO, 3240 NETDEV_UDP_TUNNEL_DROP_INFO, 3241 NETDEV_CHANGE_TX_QUEUE_LEN, 3242 NETDEV_CVLAN_FILTER_PUSH_INFO, 3243 NETDEV_CVLAN_FILTER_DROP_INFO, 3244 NETDEV_SVLAN_FILTER_PUSH_INFO, 3245 NETDEV_SVLAN_FILTER_DROP_INFO, 3246 NETDEV_OFFLOAD_XSTATS_ENABLE, 3247 NETDEV_OFFLOAD_XSTATS_DISABLE, 3248 NETDEV_OFFLOAD_XSTATS_REPORT_USED, 3249 NETDEV_OFFLOAD_XSTATS_REPORT_DELTA, 3250 NETDEV_XDP_FEAT_CHANGE, 3251 }; 3252 const char *netdev_cmd_to_name(enum netdev_cmd cmd); 3253 3254 int register_netdevice_notifier(struct notifier_block *nb); 3255 int unregister_netdevice_notifier(struct notifier_block *nb); 3256 int register_netdevice_notifier_net(struct net *net, struct notifier_block *nb); 3257 int unregister_netdevice_notifier_net(struct net *net, 3258 struct notifier_block *nb); 3259 int register_netdevice_notifier_dev_net(struct net_device *dev, 3260 struct notifier_block *nb, 3261 struct netdev_net_notifier *nn); 3262 int unregister_netdevice_notifier_dev_net(struct net_device *dev, 3263 struct notifier_block *nb, 3264 struct netdev_net_notifier *nn); 3265 3266 struct netdev_notifier_info { 3267 struct net_device *dev; 3268 struct netlink_ext_ack *extack; 3269 }; 3270 3271 struct netdev_notifier_info_ext { 3272 struct netdev_notifier_info info; /* must be first */ 3273 union { 3274 u32 mtu; 3275 } ext; 3276 }; 3277 3278 struct netdev_notifier_change_info { 3279 struct netdev_notifier_info info; /* must be first */ 3280 unsigned int flags_changed; 3281 }; 3282 3283 struct netdev_notifier_changeupper_info { 3284 struct netdev_notifier_info info; /* must be first */ 3285 struct net_device *upper_dev; /* new upper dev */ 3286 bool master; /* is upper dev master */ 3287 bool linking; /* is the notification for link or unlink */ 3288 void *upper_info; /* upper dev info */ 3289 }; 3290 3291 struct netdev_notifier_changelowerstate_info { 3292 struct netdev_notifier_info info; /* must be first */ 3293 void *lower_state_info; /* is lower dev state */ 3294 }; 3295 3296 struct netdev_notifier_pre_changeaddr_info { 3297 struct netdev_notifier_info info; /* must be first */ 3298 const unsigned char *dev_addr; 3299 }; 3300 3301 enum netdev_offload_xstats_type { 3302 NETDEV_OFFLOAD_XSTATS_TYPE_L3 = 1, 3303 }; 3304 3305 struct netdev_notifier_offload_xstats_info { 3306 struct netdev_notifier_info info; /* must be first */ 3307 enum netdev_offload_xstats_type type; 3308 3309 union { 3310 /* NETDEV_OFFLOAD_XSTATS_REPORT_DELTA */ 3311 struct netdev_notifier_offload_xstats_rd *report_delta; 3312 /* NETDEV_OFFLOAD_XSTATS_REPORT_USED */ 3313 struct netdev_notifier_offload_xstats_ru *report_used; 3314 }; 3315 }; 3316 3317 int netdev_offload_xstats_enable(struct net_device *dev, 3318 enum netdev_offload_xstats_type type, 3319 struct netlink_ext_ack *extack); 3320 int netdev_offload_xstats_disable(struct net_device *dev, 3321 enum netdev_offload_xstats_type type); 3322 bool netdev_offload_xstats_enabled(const struct net_device *dev, 3323 enum netdev_offload_xstats_type type); 3324 int netdev_offload_xstats_get(struct net_device *dev, 3325 enum netdev_offload_xstats_type type, 3326 struct rtnl_hw_stats64 *stats, bool *used, 3327 struct netlink_ext_ack *extack); 3328 void 3329 netdev_offload_xstats_report_delta(struct netdev_notifier_offload_xstats_rd *rd, 3330 const struct rtnl_hw_stats64 *stats); 3331 void 3332 netdev_offload_xstats_report_used(struct netdev_notifier_offload_xstats_ru *ru); 3333 void netdev_offload_xstats_push_delta(struct net_device *dev, 3334 enum netdev_offload_xstats_type type, 3335 const struct rtnl_hw_stats64 *stats); 3336 3337 static inline void netdev_notifier_info_init(struct netdev_notifier_info *info, 3338 struct net_device *dev) 3339 { 3340 info->dev = dev; 3341 info->extack = NULL; 3342 } 3343 3344 static inline struct net_device * 3345 netdev_notifier_info_to_dev(const struct netdev_notifier_info *info) 3346 { 3347 return info->dev; 3348 } 3349 3350 static inline struct netlink_ext_ack * 3351 netdev_notifier_info_to_extack(const struct netdev_notifier_info *info) 3352 { 3353 return info->extack; 3354 } 3355 3356 int call_netdevice_notifiers(unsigned long val, struct net_device *dev); 3357 int call_netdevice_notifiers_info(unsigned long val, 3358 struct netdev_notifier_info *info); 3359 3360 #define for_each_netdev(net, d) \ 3361 list_for_each_entry(d, &(net)->dev_base_head, dev_list) 3362 #define for_each_netdev_reverse(net, d) \ 3363 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list) 3364 #define for_each_netdev_rcu(net, d) \ 3365 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list) 3366 #define for_each_netdev_safe(net, d, n) \ 3367 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list) 3368 #define for_each_netdev_continue(net, d) \ 3369 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list) 3370 #define for_each_netdev_continue_reverse(net, d) \ 3371 list_for_each_entry_continue_reverse(d, &(net)->dev_base_head, \ 3372 dev_list) 3373 #define for_each_netdev_continue_rcu(net, d) \ 3374 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list) 3375 #define for_each_netdev_in_bond_rcu(bond, slave) \ 3376 for_each_netdev_rcu(dev_net_rcu(bond), slave) \ 3377 if (netdev_master_upper_dev_get_rcu(slave) == (bond)) 3378 #define net_device_entry(lh) list_entry(lh, struct net_device, dev_list) 3379 3380 #define for_each_netdev_dump(net, d, ifindex) \ 3381 for (; (d = xa_find(&(net)->dev_by_index, &ifindex, \ 3382 ULONG_MAX, XA_PRESENT)); ifindex++) 3383 3384 static inline struct net_device *next_net_device(struct net_device *dev) 3385 { 3386 struct list_head *lh; 3387 struct net *net; 3388 3389 net = dev_net(dev); 3390 lh = dev->dev_list.next; 3391 return lh == &net->dev_base_head ? NULL : net_device_entry(lh); 3392 } 3393 3394 static inline struct net_device *next_net_device_rcu(struct net_device *dev) 3395 { 3396 struct list_head *lh; 3397 struct net *net; 3398 3399 net = dev_net(dev); 3400 lh = rcu_dereference(list_next_rcu(&dev->dev_list)); 3401 return lh == &net->dev_base_head ? NULL : net_device_entry(lh); 3402 } 3403 3404 static inline struct net_device *first_net_device(struct net *net) 3405 { 3406 return list_empty(&net->dev_base_head) ? NULL : 3407 net_device_entry(net->dev_base_head.next); 3408 } 3409 3410 struct net_device *dev_getbyhwaddr(struct net *net, unsigned short type, 3411 const char *hwaddr); 3412 struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type, 3413 const char *hwaddr); 3414 struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type); 3415 void dev_add_pack(struct packet_type *pt); 3416 void dev_remove_pack(struct packet_type *pt); 3417 void __dev_remove_pack(struct packet_type *pt); 3418 void dev_add_offload(struct packet_offload *po); 3419 void dev_remove_offload(struct packet_offload *po); 3420 3421 int dev_get_iflink(const struct net_device *dev); 3422 int dev_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb); 3423 int dev_fill_forward_path(const struct net_device *dev, const u8 *daddr, 3424 struct net_device_path_stack *stack); 3425 struct net_device *dev_get_by_name(struct net *net, const char *name); 3426 struct net_device *dev_get_by_name_rcu(struct net *net, const char *name); 3427 struct net_device *__dev_get_by_name(struct net *net, const char *name); 3428 bool netdev_name_in_use(struct net *net, const char *name); 3429 int dev_alloc_name(struct net_device *dev, const char *name); 3430 int netif_open(struct net_device *dev, struct netlink_ext_ack *extack); 3431 int dev_open(struct net_device *dev, struct netlink_ext_ack *extack); 3432 void netif_close(struct net_device *dev); 3433 void dev_close(struct net_device *dev); 3434 void netif_close_many(struct list_head *head, bool unlink); 3435 void netif_disable_lro(struct net_device *dev); 3436 void dev_disable_lro(struct net_device *dev); 3437 int dev_loopback_xmit(struct net *net, struct sock *sk, struct sk_buff *newskb); 3438 u16 dev_pick_tx_zero(struct net_device *dev, struct sk_buff *skb, 3439 struct net_device *sb_dev); 3440 3441 int __dev_queue_xmit(struct sk_buff *skb, struct net_device *sb_dev); 3442 int __dev_direct_xmit(struct sk_buff *skb, u16 queue_id); 3443 3444 static inline int dev_queue_xmit(struct sk_buff *skb) 3445 { 3446 return __dev_queue_xmit(skb, NULL); 3447 } 3448 3449 static inline int dev_queue_xmit_accel(struct sk_buff *skb, 3450 struct net_device *sb_dev) 3451 { 3452 return __dev_queue_xmit(skb, sb_dev); 3453 } 3454 3455 static inline int dev_direct_xmit(struct sk_buff *skb, u16 queue_id) 3456 { 3457 int ret; 3458 3459 ret = __dev_direct_xmit(skb, queue_id); 3460 if (!dev_xmit_complete(ret)) 3461 kfree_skb(skb); 3462 return ret; 3463 } 3464 3465 int register_netdevice(struct net_device *dev); 3466 void unregister_netdevice_queue(struct net_device *dev, struct list_head *head); 3467 void unregister_netdevice_many(struct list_head *head); 3468 bool unregister_netdevice_queued(const struct net_device *dev); 3469 3470 static inline void unregister_netdevice(struct net_device *dev) 3471 { 3472 unregister_netdevice_queue(dev, NULL); 3473 } 3474 3475 int netdev_refcnt_read(const struct net_device *dev); 3476 void free_netdev(struct net_device *dev); 3477 3478 struct net_device *netdev_get_xmit_slave(struct net_device *dev, 3479 struct sk_buff *skb, 3480 bool all_slaves); 3481 struct net_device *netdev_sk_get_lowest_dev(struct net_device *dev, 3482 struct sock *sk); 3483 struct net_device *dev_get_by_index(struct net *net, int ifindex); 3484 struct net_device *__dev_get_by_index(struct net *net, int ifindex); 3485 struct net_device *netdev_get_by_index(struct net *net, int ifindex, 3486 netdevice_tracker *tracker, gfp_t gfp); 3487 struct net_device *netdev_get_by_index_lock(struct net *net, int ifindex); 3488 struct net_device *netdev_get_by_name(struct net *net, const char *name, 3489 netdevice_tracker *tracker, gfp_t gfp); 3490 struct net_device *netdev_get_by_flags_rcu(struct net *net, netdevice_tracker *tracker, 3491 unsigned short flags, unsigned short mask); 3492 struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex); 3493 void netdev_copy_name(struct net_device *dev, char *name); 3494 3495 static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev, 3496 unsigned short type, 3497 const void *daddr, const void *saddr, 3498 unsigned int len) 3499 { 3500 if (!dev->header_ops || !dev->header_ops->create) 3501 return 0; 3502 3503 return dev->header_ops->create(skb, dev, type, daddr, saddr, len); 3504 } 3505 3506 static inline int dev_parse_header(const struct sk_buff *skb, 3507 unsigned char *haddr) 3508 { 3509 const struct net_device *dev = skb->dev; 3510 3511 if (!dev->header_ops || !dev->header_ops->parse) 3512 return 0; 3513 return dev->header_ops->parse(skb, dev, haddr); 3514 } 3515 3516 static inline __be16 dev_parse_header_protocol(const struct sk_buff *skb) 3517 { 3518 const struct net_device *dev = skb->dev; 3519 3520 if (!dev->header_ops || !dev->header_ops->parse_protocol) 3521 return 0; 3522 return dev->header_ops->parse_protocol(skb); 3523 } 3524 3525 /* ll_header must have at least hard_header_len allocated */ 3526 static inline bool dev_validate_header(const struct net_device *dev, 3527 char *ll_header, int len) 3528 { 3529 if (likely(len >= dev->hard_header_len)) 3530 return true; 3531 if (len < dev->min_header_len) 3532 return false; 3533 3534 if (capable(CAP_SYS_RAWIO)) { 3535 memset(ll_header + len, 0, dev->hard_header_len - len); 3536 return true; 3537 } 3538 3539 if (dev->header_ops && dev->header_ops->validate) 3540 return dev->header_ops->validate(ll_header, len); 3541 3542 return false; 3543 } 3544 3545 static inline bool dev_has_header(const struct net_device *dev) 3546 { 3547 return dev->header_ops && dev->header_ops->create; 3548 } 3549 3550 struct numa_drop_counters { 3551 atomic_t drops0 ____cacheline_aligned_in_smp; 3552 atomic_t drops1 ____cacheline_aligned_in_smp; 3553 }; 3554 3555 static inline int numa_drop_read(const struct numa_drop_counters *ndc) 3556 { 3557 return atomic_read(&ndc->drops0) + atomic_read(&ndc->drops1); 3558 } 3559 3560 static inline void numa_drop_add(struct numa_drop_counters *ndc, int val) 3561 { 3562 int n = numa_node_id() % 2; 3563 3564 if (n) 3565 atomic_add(val, &ndc->drops1); 3566 else 3567 atomic_add(val, &ndc->drops0); 3568 } 3569 3570 static inline void numa_drop_reset(struct numa_drop_counters *ndc) 3571 { 3572 atomic_set(&ndc->drops0, 0); 3573 atomic_set(&ndc->drops1, 0); 3574 } 3575 3576 /* 3577 * Incoming packets are placed on per-CPU queues 3578 */ 3579 struct softnet_data { 3580 struct list_head poll_list; 3581 struct sk_buff_head process_queue; 3582 local_lock_t process_queue_bh_lock; 3583 3584 /* stats */ 3585 unsigned int processed; 3586 unsigned int time_squeeze; 3587 #ifdef CONFIG_RPS 3588 struct softnet_data *rps_ipi_list; 3589 #endif 3590 3591 unsigned int received_rps; 3592 bool in_net_rx_action; 3593 bool in_napi_threaded_poll; 3594 3595 #ifdef CONFIG_NET_FLOW_LIMIT 3596 struct sd_flow_limit __rcu *flow_limit; 3597 #endif 3598 struct Qdisc *output_queue; 3599 struct Qdisc **output_queue_tailp; 3600 struct sk_buff *completion_queue; 3601 #ifdef CONFIG_XFRM_OFFLOAD 3602 struct sk_buff_head xfrm_backlog; 3603 #endif 3604 /* written and read only by owning cpu: */ 3605 struct netdev_xmit xmit; 3606 #ifdef CONFIG_RPS 3607 /* input_queue_head should be written by cpu owning this struct, 3608 * and only read by other cpus. Worth using a cache line. 3609 */ 3610 unsigned int input_queue_head ____cacheline_aligned_in_smp; 3611 3612 /* Elements below can be accessed between CPUs for RPS/RFS */ 3613 call_single_data_t csd ____cacheline_aligned_in_smp; 3614 struct softnet_data *rps_ipi_next; 3615 unsigned int cpu; 3616 3617 /* We force a cacheline alignment from here, to hold together 3618 * input_queue_tail, input_pkt_queue and backlog.state. 3619 * We add holes so that backlog.state is the last field 3620 * of this cache line. 3621 */ 3622 long pad[3] ____cacheline_aligned_in_smp; 3623 unsigned int input_queue_tail; 3624 #endif 3625 struct sk_buff_head input_pkt_queue; 3626 3627 struct napi_struct backlog; 3628 3629 struct numa_drop_counters drop_counters; 3630 3631 int defer_ipi_scheduled ____cacheline_aligned_in_smp; 3632 call_single_data_t defer_csd; 3633 }; 3634 3635 DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data); 3636 3637 struct page_pool_bh { 3638 struct page_pool *pool; 3639 local_lock_t bh_lock; 3640 }; 3641 DECLARE_PER_CPU(struct page_pool_bh, system_page_pool); 3642 3643 #define XMIT_RECURSION_LIMIT 8 3644 3645 #ifndef CONFIG_PREEMPT_RT 3646 static inline int dev_recursion_level(void) 3647 { 3648 return this_cpu_read(softnet_data.xmit.recursion); 3649 } 3650 3651 static inline bool dev_xmit_recursion(void) 3652 { 3653 return unlikely(__this_cpu_read(softnet_data.xmit.recursion) > 3654 XMIT_RECURSION_LIMIT); 3655 } 3656 3657 static inline void dev_xmit_recursion_inc(void) 3658 { 3659 __this_cpu_inc(softnet_data.xmit.recursion); 3660 } 3661 3662 static inline void dev_xmit_recursion_dec(void) 3663 { 3664 __this_cpu_dec(softnet_data.xmit.recursion); 3665 } 3666 #else 3667 static inline int dev_recursion_level(void) 3668 { 3669 return current->net_xmit.recursion; 3670 } 3671 3672 static inline bool dev_xmit_recursion(void) 3673 { 3674 return unlikely(current->net_xmit.recursion > XMIT_RECURSION_LIMIT); 3675 } 3676 3677 static inline void dev_xmit_recursion_inc(void) 3678 { 3679 current->net_xmit.recursion++; 3680 } 3681 3682 static inline void dev_xmit_recursion_dec(void) 3683 { 3684 current->net_xmit.recursion--; 3685 } 3686 #endif 3687 3688 void __netif_schedule(struct Qdisc *q); 3689 void netif_schedule_queue(struct netdev_queue *txq); 3690 3691 static inline void netif_tx_schedule_all(struct net_device *dev) 3692 { 3693 unsigned int i; 3694 3695 for (i = 0; i < dev->num_tx_queues; i++) 3696 netif_schedule_queue(netdev_get_tx_queue(dev, i)); 3697 } 3698 3699 static __always_inline void netif_tx_start_queue(struct netdev_queue *dev_queue) 3700 { 3701 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state); 3702 } 3703 3704 /** 3705 * netif_start_queue - allow transmit 3706 * @dev: network device 3707 * 3708 * Allow upper layers to call the device hard_start_xmit routine. 3709 */ 3710 static inline void netif_start_queue(struct net_device *dev) 3711 { 3712 netif_tx_start_queue(netdev_get_tx_queue(dev, 0)); 3713 } 3714 3715 static inline void netif_tx_start_all_queues(struct net_device *dev) 3716 { 3717 unsigned int i; 3718 3719 for (i = 0; i < dev->num_tx_queues; i++) { 3720 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 3721 netif_tx_start_queue(txq); 3722 } 3723 } 3724 3725 void netif_tx_wake_queue(struct netdev_queue *dev_queue); 3726 3727 /** 3728 * netif_wake_queue - restart transmit 3729 * @dev: network device 3730 * 3731 * Allow upper layers to call the device hard_start_xmit routine. 3732 * Used for flow control when transmit resources are available. 3733 */ 3734 static inline void netif_wake_queue(struct net_device *dev) 3735 { 3736 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0)); 3737 } 3738 3739 static inline void netif_tx_wake_all_queues(struct net_device *dev) 3740 { 3741 unsigned int i; 3742 3743 for (i = 0; i < dev->num_tx_queues; i++) { 3744 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 3745 netif_tx_wake_queue(txq); 3746 } 3747 } 3748 3749 static __always_inline void netif_tx_stop_queue(struct netdev_queue *dev_queue) 3750 { 3751 /* Paired with READ_ONCE() from dev_watchdog() */ 3752 WRITE_ONCE(dev_queue->trans_start, jiffies); 3753 3754 /* This barrier is paired with smp_mb() from dev_watchdog() */ 3755 smp_mb__before_atomic(); 3756 3757 /* Must be an atomic op see netif_txq_try_stop() */ 3758 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state); 3759 } 3760 3761 /** 3762 * netif_stop_queue - stop transmitted packets 3763 * @dev: network device 3764 * 3765 * Stop upper layers calling the device hard_start_xmit routine. 3766 * Used for flow control when transmit resources are unavailable. 3767 */ 3768 static inline void netif_stop_queue(struct net_device *dev) 3769 { 3770 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0)); 3771 } 3772 3773 void netif_tx_stop_all_queues(struct net_device *dev); 3774 3775 static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue) 3776 { 3777 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state); 3778 } 3779 3780 /** 3781 * netif_queue_stopped - test if transmit queue is flowblocked 3782 * @dev: network device 3783 * 3784 * Test if transmit queue on device is currently unable to send. 3785 */ 3786 static inline bool netif_queue_stopped(const struct net_device *dev) 3787 { 3788 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0)); 3789 } 3790 3791 static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue) 3792 { 3793 return dev_queue->state & QUEUE_STATE_ANY_XOFF; 3794 } 3795 3796 static inline bool 3797 netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue) 3798 { 3799 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN; 3800 } 3801 3802 static inline bool 3803 netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue) 3804 { 3805 return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN; 3806 } 3807 3808 /** 3809 * netdev_queue_set_dql_min_limit - set dql minimum limit 3810 * @dev_queue: pointer to transmit queue 3811 * @min_limit: dql minimum limit 3812 * 3813 * Forces xmit_more() to return true until the minimum threshold 3814 * defined by @min_limit is reached (or until the tx queue is 3815 * empty). Warning: to be use with care, misuse will impact the 3816 * latency. 3817 */ 3818 static inline void netdev_queue_set_dql_min_limit(struct netdev_queue *dev_queue, 3819 unsigned int min_limit) 3820 { 3821 #ifdef CONFIG_BQL 3822 dev_queue->dql.min_limit = min_limit; 3823 #endif 3824 } 3825 3826 static inline int netdev_queue_dql_avail(const struct netdev_queue *txq) 3827 { 3828 #ifdef CONFIG_BQL 3829 /* Non-BQL migrated drivers will return 0, too. */ 3830 return dql_avail(&txq->dql); 3831 #else 3832 return 0; 3833 #endif 3834 } 3835 3836 /** 3837 * netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write 3838 * @dev_queue: pointer to transmit queue 3839 * 3840 * BQL enabled drivers might use this helper in their ndo_start_xmit(), 3841 * to give appropriate hint to the CPU. 3842 */ 3843 static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue) 3844 { 3845 #ifdef CONFIG_BQL 3846 prefetchw(&dev_queue->dql.num_queued); 3847 #endif 3848 } 3849 3850 /** 3851 * netdev_txq_bql_complete_prefetchw - prefetch bql data for write 3852 * @dev_queue: pointer to transmit queue 3853 * 3854 * BQL enabled drivers might use this helper in their TX completion path, 3855 * to give appropriate hint to the CPU. 3856 */ 3857 static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue) 3858 { 3859 #ifdef CONFIG_BQL 3860 prefetchw(&dev_queue->dql.limit); 3861 #endif 3862 } 3863 3864 /** 3865 * netdev_tx_sent_queue - report the number of bytes queued to a given tx queue 3866 * @dev_queue: network device queue 3867 * @bytes: number of bytes queued to the device queue 3868 * 3869 * Report the number of bytes queued for sending/completion to the network 3870 * device hardware queue. @bytes should be a good approximation and should 3871 * exactly match netdev_completed_queue() @bytes. 3872 * This is typically called once per packet, from ndo_start_xmit(). 3873 */ 3874 static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue, 3875 unsigned int bytes) 3876 { 3877 #ifdef CONFIG_BQL 3878 dql_queued(&dev_queue->dql, bytes); 3879 3880 if (likely(dql_avail(&dev_queue->dql) >= 0)) 3881 return; 3882 3883 /* Paired with READ_ONCE() from dev_watchdog() */ 3884 WRITE_ONCE(dev_queue->trans_start, jiffies); 3885 3886 /* This barrier is paired with smp_mb() from dev_watchdog() */ 3887 smp_mb__before_atomic(); 3888 3889 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state); 3890 3891 /* 3892 * The XOFF flag must be set before checking the dql_avail below, 3893 * because in netdev_tx_completed_queue we update the dql_completed 3894 * before checking the XOFF flag. 3895 */ 3896 smp_mb__after_atomic(); 3897 3898 /* check again in case another CPU has just made room avail */ 3899 if (unlikely(dql_avail(&dev_queue->dql) >= 0)) 3900 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state); 3901 #endif 3902 } 3903 3904 /* Variant of netdev_tx_sent_queue() for drivers that are aware 3905 * that they should not test BQL status themselves. 3906 * We do want to change __QUEUE_STATE_STACK_XOFF only for the last 3907 * skb of a batch. 3908 * Returns true if the doorbell must be used to kick the NIC. 3909 */ 3910 static inline bool __netdev_tx_sent_queue(struct netdev_queue *dev_queue, 3911 unsigned int bytes, 3912 bool xmit_more) 3913 { 3914 if (xmit_more) { 3915 #ifdef CONFIG_BQL 3916 dql_queued(&dev_queue->dql, bytes); 3917 #endif 3918 return netif_tx_queue_stopped(dev_queue); 3919 } 3920 netdev_tx_sent_queue(dev_queue, bytes); 3921 return true; 3922 } 3923 3924 /** 3925 * netdev_sent_queue - report the number of bytes queued to hardware 3926 * @dev: network device 3927 * @bytes: number of bytes queued to the hardware device queue 3928 * 3929 * Report the number of bytes queued for sending/completion to the network 3930 * device hardware queue#0. @bytes should be a good approximation and should 3931 * exactly match netdev_completed_queue() @bytes. 3932 * This is typically called once per packet, from ndo_start_xmit(). 3933 */ 3934 static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes) 3935 { 3936 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes); 3937 } 3938 3939 static inline bool __netdev_sent_queue(struct net_device *dev, 3940 unsigned int bytes, 3941 bool xmit_more) 3942 { 3943 return __netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes, 3944 xmit_more); 3945 } 3946 3947 /** 3948 * netdev_tx_completed_queue - report number of packets/bytes at TX completion. 3949 * @dev_queue: network device queue 3950 * @pkts: number of packets (currently ignored) 3951 * @bytes: number of bytes dequeued from the device queue 3952 * 3953 * Must be called at most once per TX completion round (and not per 3954 * individual packet), so that BQL can adjust its limits appropriately. 3955 */ 3956 static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue, 3957 unsigned int pkts, unsigned int bytes) 3958 { 3959 #ifdef CONFIG_BQL 3960 if (unlikely(!bytes)) 3961 return; 3962 3963 dql_completed(&dev_queue->dql, bytes); 3964 3965 /* 3966 * Without the memory barrier there is a small possibility that 3967 * netdev_tx_sent_queue will miss the update and cause the queue to 3968 * be stopped forever 3969 */ 3970 smp_mb(); /* NOTE: netdev_txq_completed_mb() assumes this exists */ 3971 3972 if (unlikely(dql_avail(&dev_queue->dql) < 0)) 3973 return; 3974 3975 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state)) 3976 netif_schedule_queue(dev_queue); 3977 #endif 3978 } 3979 3980 /** 3981 * netdev_completed_queue - report bytes and packets completed by device 3982 * @dev: network device 3983 * @pkts: actual number of packets sent over the medium 3984 * @bytes: actual number of bytes sent over the medium 3985 * 3986 * Report the number of bytes and packets transmitted by the network device 3987 * hardware queue over the physical medium, @bytes must exactly match the 3988 * @bytes amount passed to netdev_sent_queue() 3989 */ 3990 static inline void netdev_completed_queue(struct net_device *dev, 3991 unsigned int pkts, unsigned int bytes) 3992 { 3993 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes); 3994 } 3995 3996 static inline void netdev_tx_reset_queue(struct netdev_queue *q) 3997 { 3998 #ifdef CONFIG_BQL 3999 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state); 4000 dql_reset(&q->dql); 4001 #endif 4002 } 4003 4004 /** 4005 * netdev_tx_reset_subqueue - reset the BQL stats and state of a netdev queue 4006 * @dev: network device 4007 * @qid: stack index of the queue to reset 4008 */ 4009 static inline void netdev_tx_reset_subqueue(const struct net_device *dev, 4010 u32 qid) 4011 { 4012 netdev_tx_reset_queue(netdev_get_tx_queue(dev, qid)); 4013 } 4014 4015 /** 4016 * netdev_reset_queue - reset the packets and bytes count of a network device 4017 * @dev_queue: network device 4018 * 4019 * Reset the bytes and packet count of a network device and clear the 4020 * software flow control OFF bit for this network device 4021 */ 4022 static inline void netdev_reset_queue(struct net_device *dev_queue) 4023 { 4024 netdev_tx_reset_subqueue(dev_queue, 0); 4025 } 4026 4027 /** 4028 * netdev_cap_txqueue - check if selected tx queue exceeds device queues 4029 * @dev: network device 4030 * @queue_index: given tx queue index 4031 * 4032 * Returns 0 if given tx queue index >= number of device tx queues, 4033 * otherwise returns the originally passed tx queue index. 4034 */ 4035 static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index) 4036 { 4037 if (unlikely(queue_index >= dev->real_num_tx_queues)) { 4038 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n", 4039 dev->name, queue_index, 4040 dev->real_num_tx_queues); 4041 return 0; 4042 } 4043 4044 return queue_index; 4045 } 4046 4047 /** 4048 * netif_running - test if up 4049 * @dev: network device 4050 * 4051 * Test if the device has been brought up. 4052 */ 4053 static inline bool netif_running(const struct net_device *dev) 4054 { 4055 return test_bit(__LINK_STATE_START, &dev->state); 4056 } 4057 4058 /* 4059 * Routines to manage the subqueues on a device. We only need start, 4060 * stop, and a check if it's stopped. All other device management is 4061 * done at the overall netdevice level. 4062 * Also test the device if we're multiqueue. 4063 */ 4064 4065 /** 4066 * netif_start_subqueue - allow sending packets on subqueue 4067 * @dev: network device 4068 * @queue_index: sub queue index 4069 * 4070 * Start individual transmit queue of a device with multiple transmit queues. 4071 */ 4072 static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index) 4073 { 4074 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 4075 4076 netif_tx_start_queue(txq); 4077 } 4078 4079 /** 4080 * netif_stop_subqueue - stop sending packets on subqueue 4081 * @dev: network device 4082 * @queue_index: sub queue index 4083 * 4084 * Stop individual transmit queue of a device with multiple transmit queues. 4085 */ 4086 static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index) 4087 { 4088 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 4089 netif_tx_stop_queue(txq); 4090 } 4091 4092 /** 4093 * __netif_subqueue_stopped - test status of subqueue 4094 * @dev: network device 4095 * @queue_index: sub queue index 4096 * 4097 * Check individual transmit queue of a device with multiple transmit queues. 4098 */ 4099 static inline bool __netif_subqueue_stopped(const struct net_device *dev, 4100 u16 queue_index) 4101 { 4102 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 4103 4104 return netif_tx_queue_stopped(txq); 4105 } 4106 4107 /** 4108 * netif_subqueue_stopped - test status of subqueue 4109 * @dev: network device 4110 * @skb: sub queue buffer pointer 4111 * 4112 * Check individual transmit queue of a device with multiple transmit queues. 4113 */ 4114 static inline bool netif_subqueue_stopped(const struct net_device *dev, 4115 struct sk_buff *skb) 4116 { 4117 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb)); 4118 } 4119 4120 /** 4121 * netif_wake_subqueue - allow sending packets on subqueue 4122 * @dev: network device 4123 * @queue_index: sub queue index 4124 * 4125 * Resume individual transmit queue of a device with multiple transmit queues. 4126 */ 4127 static inline void netif_wake_subqueue(struct net_device *dev, u16 queue_index) 4128 { 4129 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 4130 4131 netif_tx_wake_queue(txq); 4132 } 4133 4134 #ifdef CONFIG_XPS 4135 int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask, 4136 u16 index); 4137 int __netif_set_xps_queue(struct net_device *dev, const unsigned long *mask, 4138 u16 index, enum xps_map_type type); 4139 4140 /** 4141 * netif_attr_test_mask - Test a CPU or Rx queue set in a mask 4142 * @j: CPU/Rx queue index 4143 * @mask: bitmask of all cpus/rx queues 4144 * @nr_bits: number of bits in the bitmask 4145 * 4146 * Test if a CPU or Rx queue index is set in a mask of all CPU/Rx queues. 4147 */ 4148 static inline bool netif_attr_test_mask(unsigned long j, 4149 const unsigned long *mask, 4150 unsigned int nr_bits) 4151 { 4152 cpu_max_bits_warn(j, nr_bits); 4153 return test_bit(j, mask); 4154 } 4155 4156 /** 4157 * netif_attr_test_online - Test for online CPU/Rx queue 4158 * @j: CPU/Rx queue index 4159 * @online_mask: bitmask for CPUs/Rx queues that are online 4160 * @nr_bits: number of bits in the bitmask 4161 * 4162 * Returns: true if a CPU/Rx queue is online. 4163 */ 4164 static inline bool netif_attr_test_online(unsigned long j, 4165 const unsigned long *online_mask, 4166 unsigned int nr_bits) 4167 { 4168 cpu_max_bits_warn(j, nr_bits); 4169 4170 if (online_mask) 4171 return test_bit(j, online_mask); 4172 4173 return (j < nr_bits); 4174 } 4175 4176 /** 4177 * netif_attrmask_next - get the next CPU/Rx queue in a cpu/Rx queues mask 4178 * @n: CPU/Rx queue index 4179 * @srcp: the cpumask/Rx queue mask pointer 4180 * @nr_bits: number of bits in the bitmask 4181 * 4182 * Returns: next (after n) CPU/Rx queue index in the mask; 4183 * >= nr_bits if no further CPUs/Rx queues set. 4184 */ 4185 static inline unsigned int netif_attrmask_next(int n, const unsigned long *srcp, 4186 unsigned int nr_bits) 4187 { 4188 /* -1 is a legal arg here. */ 4189 if (n != -1) 4190 cpu_max_bits_warn(n, nr_bits); 4191 4192 if (srcp) 4193 return find_next_bit(srcp, nr_bits, n + 1); 4194 4195 return n + 1; 4196 } 4197 4198 /** 4199 * netif_attrmask_next_and - get the next CPU/Rx queue in \*src1p & \*src2p 4200 * @n: CPU/Rx queue index 4201 * @src1p: the first CPUs/Rx queues mask pointer 4202 * @src2p: the second CPUs/Rx queues mask pointer 4203 * @nr_bits: number of bits in the bitmask 4204 * 4205 * Returns: next (after n) CPU/Rx queue index set in both masks; 4206 * >= nr_bits if no further CPUs/Rx queues set in both. 4207 */ 4208 static inline int netif_attrmask_next_and(int n, const unsigned long *src1p, 4209 const unsigned long *src2p, 4210 unsigned int nr_bits) 4211 { 4212 /* -1 is a legal arg here. */ 4213 if (n != -1) 4214 cpu_max_bits_warn(n, nr_bits); 4215 4216 if (src1p && src2p) 4217 return find_next_and_bit(src1p, src2p, nr_bits, n + 1); 4218 else if (src1p) 4219 return find_next_bit(src1p, nr_bits, n + 1); 4220 else if (src2p) 4221 return find_next_bit(src2p, nr_bits, n + 1); 4222 4223 return n + 1; 4224 } 4225 #else 4226 static inline int netif_set_xps_queue(struct net_device *dev, 4227 const struct cpumask *mask, 4228 u16 index) 4229 { 4230 return 0; 4231 } 4232 4233 static inline int __netif_set_xps_queue(struct net_device *dev, 4234 const unsigned long *mask, 4235 u16 index, enum xps_map_type type) 4236 { 4237 return 0; 4238 } 4239 #endif 4240 4241 /** 4242 * netif_is_multiqueue - test if device has multiple transmit queues 4243 * @dev: network device 4244 * 4245 * Check if device has multiple transmit queues 4246 */ 4247 static inline bool netif_is_multiqueue(const struct net_device *dev) 4248 { 4249 return dev->num_tx_queues > 1; 4250 } 4251 4252 int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq); 4253 int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq); 4254 int netif_set_real_num_queues(struct net_device *dev, 4255 unsigned int txq, unsigned int rxq); 4256 4257 int netif_get_num_default_rss_queues(void); 4258 4259 void dev_kfree_skb_irq_reason(struct sk_buff *skb, enum skb_drop_reason reason); 4260 void dev_kfree_skb_any_reason(struct sk_buff *skb, enum skb_drop_reason reason); 4261 4262 /* 4263 * It is not allowed to call kfree_skb() or consume_skb() from hardware 4264 * interrupt context or with hardware interrupts being disabled. 4265 * (in_hardirq() || irqs_disabled()) 4266 * 4267 * We provide four helpers that can be used in following contexts : 4268 * 4269 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context, 4270 * replacing kfree_skb(skb) 4271 * 4272 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context. 4273 * Typically used in place of consume_skb(skb) in TX completion path 4274 * 4275 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context, 4276 * replacing kfree_skb(skb) 4277 * 4278 * dev_consume_skb_any(skb) when caller doesn't know its current irq context, 4279 * and consumed a packet. Used in place of consume_skb(skb) 4280 */ 4281 static inline void dev_kfree_skb_irq(struct sk_buff *skb) 4282 { 4283 dev_kfree_skb_irq_reason(skb, SKB_DROP_REASON_NOT_SPECIFIED); 4284 } 4285 4286 static inline void dev_consume_skb_irq(struct sk_buff *skb) 4287 { 4288 dev_kfree_skb_irq_reason(skb, SKB_CONSUMED); 4289 } 4290 4291 static inline void dev_kfree_skb_any(struct sk_buff *skb) 4292 { 4293 dev_kfree_skb_any_reason(skb, SKB_DROP_REASON_NOT_SPECIFIED); 4294 } 4295 4296 static inline void dev_consume_skb_any(struct sk_buff *skb) 4297 { 4298 dev_kfree_skb_any_reason(skb, SKB_CONSUMED); 4299 } 4300 4301 u32 bpf_prog_run_generic_xdp(struct sk_buff *skb, struct xdp_buff *xdp, 4302 const struct bpf_prog *xdp_prog); 4303 void generic_xdp_tx(struct sk_buff *skb, const struct bpf_prog *xdp_prog); 4304 int do_xdp_generic(const struct bpf_prog *xdp_prog, struct sk_buff **pskb); 4305 int netif_rx(struct sk_buff *skb); 4306 int __netif_rx(struct sk_buff *skb); 4307 4308 int netif_receive_skb(struct sk_buff *skb); 4309 int netif_receive_skb_core(struct sk_buff *skb); 4310 void netif_receive_skb_list_internal(struct list_head *head); 4311 void netif_receive_skb_list(struct list_head *head); 4312 gro_result_t gro_receive_skb(struct gro_node *gro, struct sk_buff *skb); 4313 4314 static inline gro_result_t napi_gro_receive(struct napi_struct *napi, 4315 struct sk_buff *skb) 4316 { 4317 return gro_receive_skb(&napi->gro, skb); 4318 } 4319 4320 struct sk_buff *napi_get_frags(struct napi_struct *napi); 4321 gro_result_t napi_gro_frags(struct napi_struct *napi); 4322 4323 static inline void napi_free_frags(struct napi_struct *napi) 4324 { 4325 kfree_skb(napi->skb); 4326 napi->skb = NULL; 4327 } 4328 4329 bool netdev_is_rx_handler_busy(struct net_device *dev); 4330 int netdev_rx_handler_register(struct net_device *dev, 4331 rx_handler_func_t *rx_handler, 4332 void *rx_handler_data); 4333 void netdev_rx_handler_unregister(struct net_device *dev); 4334 4335 bool dev_valid_name(const char *name); 4336 static inline bool is_socket_ioctl_cmd(unsigned int cmd) 4337 { 4338 return _IOC_TYPE(cmd) == SOCK_IOC_TYPE; 4339 } 4340 int get_user_ifreq(struct ifreq *ifr, void __user **ifrdata, void __user *arg); 4341 int put_user_ifreq(struct ifreq *ifr, void __user *arg); 4342 int dev_ioctl(struct net *net, unsigned int cmd, struct ifreq *ifr, 4343 void __user *data, bool *need_copyout); 4344 int dev_ifconf(struct net *net, struct ifconf __user *ifc); 4345 int dev_eth_ioctl(struct net_device *dev, 4346 struct ifreq *ifr, unsigned int cmd); 4347 int generic_hwtstamp_get_lower(struct net_device *dev, 4348 struct kernel_hwtstamp_config *kernel_cfg); 4349 int generic_hwtstamp_set_lower(struct net_device *dev, 4350 struct kernel_hwtstamp_config *kernel_cfg, 4351 struct netlink_ext_ack *extack); 4352 int dev_ethtool(struct net *net, struct ifreq *ifr, void __user *userdata); 4353 unsigned int netif_get_flags(const struct net_device *dev); 4354 int __dev_change_flags(struct net_device *dev, unsigned int flags, 4355 struct netlink_ext_ack *extack); 4356 int netif_change_flags(struct net_device *dev, unsigned int flags, 4357 struct netlink_ext_ack *extack); 4358 int dev_change_flags(struct net_device *dev, unsigned int flags, 4359 struct netlink_ext_ack *extack); 4360 int netif_set_alias(struct net_device *dev, const char *alias, size_t len); 4361 int dev_set_alias(struct net_device *, const char *, size_t); 4362 int dev_get_alias(const struct net_device *, char *, size_t); 4363 int __dev_change_net_namespace(struct net_device *dev, struct net *net, 4364 const char *pat, int new_ifindex, 4365 struct netlink_ext_ack *extack); 4366 int dev_change_net_namespace(struct net_device *dev, struct net *net, 4367 const char *pat); 4368 int __netif_set_mtu(struct net_device *dev, int new_mtu); 4369 int netif_set_mtu(struct net_device *dev, int new_mtu); 4370 int dev_set_mtu(struct net_device *, int); 4371 int netif_pre_changeaddr_notify(struct net_device *dev, const char *addr, 4372 struct netlink_ext_ack *extack); 4373 int netif_set_mac_address(struct net_device *dev, struct sockaddr_storage *ss, 4374 struct netlink_ext_ack *extack); 4375 int dev_set_mac_address(struct net_device *dev, struct sockaddr_storage *ss, 4376 struct netlink_ext_ack *extack); 4377 int dev_set_mac_address_user(struct net_device *dev, struct sockaddr_storage *ss, 4378 struct netlink_ext_ack *extack); 4379 int netif_get_mac_address(struct sockaddr *sa, struct net *net, char *dev_name); 4380 int netif_get_port_parent_id(struct net_device *dev, 4381 struct netdev_phys_item_id *ppid, bool recurse); 4382 bool netdev_port_same_parent_id(struct net_device *a, struct net_device *b); 4383 4384 struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev, bool *again); 4385 struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev, 4386 struct netdev_queue *txq, int *ret); 4387 4388 int bpf_xdp_link_attach(const union bpf_attr *attr, struct bpf_prog *prog); 4389 u8 dev_xdp_prog_count(struct net_device *dev); 4390 int netif_xdp_propagate(struct net_device *dev, struct netdev_bpf *bpf); 4391 int dev_xdp_propagate(struct net_device *dev, struct netdev_bpf *bpf); 4392 u8 dev_xdp_sb_prog_count(struct net_device *dev); 4393 u32 dev_xdp_prog_id(struct net_device *dev, enum bpf_xdp_mode mode); 4394 4395 u32 dev_get_min_mp_channel_count(const struct net_device *dev); 4396 4397 int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb); 4398 int dev_forward_skb(struct net_device *dev, struct sk_buff *skb); 4399 int dev_forward_skb_nomtu(struct net_device *dev, struct sk_buff *skb); 4400 bool is_skb_forwardable(const struct net_device *dev, 4401 const struct sk_buff *skb); 4402 4403 static __always_inline bool __is_skb_forwardable(const struct net_device *dev, 4404 const struct sk_buff *skb, 4405 const bool check_mtu) 4406 { 4407 const u32 vlan_hdr_len = 4; /* VLAN_HLEN */ 4408 unsigned int len; 4409 4410 if (!(dev->flags & IFF_UP)) 4411 return false; 4412 4413 if (!check_mtu) 4414 return true; 4415 4416 len = dev->mtu + dev->hard_header_len + vlan_hdr_len; 4417 if (skb->len <= len) 4418 return true; 4419 4420 /* if TSO is enabled, we don't care about the length as the packet 4421 * could be forwarded without being segmented before 4422 */ 4423 if (skb_is_gso(skb)) 4424 return true; 4425 4426 return false; 4427 } 4428 4429 void netdev_core_stats_inc(struct net_device *dev, u32 offset); 4430 4431 #define DEV_CORE_STATS_INC(FIELD) \ 4432 static inline void dev_core_stats_##FIELD##_inc(struct net_device *dev) \ 4433 { \ 4434 netdev_core_stats_inc(dev, \ 4435 offsetof(struct net_device_core_stats, FIELD)); \ 4436 } 4437 DEV_CORE_STATS_INC(rx_dropped) 4438 DEV_CORE_STATS_INC(tx_dropped) 4439 DEV_CORE_STATS_INC(rx_nohandler) 4440 DEV_CORE_STATS_INC(rx_otherhost_dropped) 4441 #undef DEV_CORE_STATS_INC 4442 4443 static __always_inline int ____dev_forward_skb(struct net_device *dev, 4444 struct sk_buff *skb, 4445 const bool check_mtu) 4446 { 4447 if (skb_orphan_frags(skb, GFP_ATOMIC) || 4448 unlikely(!__is_skb_forwardable(dev, skb, check_mtu))) { 4449 dev_core_stats_rx_dropped_inc(dev); 4450 kfree_skb(skb); 4451 return NET_RX_DROP; 4452 } 4453 4454 skb_scrub_packet(skb, !net_eq(dev_net(dev), dev_net(skb->dev))); 4455 skb->priority = 0; 4456 return 0; 4457 } 4458 4459 bool dev_nit_active_rcu(const struct net_device *dev); 4460 static inline bool dev_nit_active(const struct net_device *dev) 4461 { 4462 bool ret; 4463 4464 rcu_read_lock(); 4465 ret = dev_nit_active_rcu(dev); 4466 rcu_read_unlock(); 4467 return ret; 4468 } 4469 4470 void dev_queue_xmit_nit(struct sk_buff *skb, struct net_device *dev); 4471 4472 static inline void __dev_put(struct net_device *dev) 4473 { 4474 if (dev) { 4475 #ifdef CONFIG_PCPU_DEV_REFCNT 4476 this_cpu_dec(*dev->pcpu_refcnt); 4477 #else 4478 refcount_dec(&dev->dev_refcnt); 4479 #endif 4480 } 4481 } 4482 4483 static inline void __dev_hold(struct net_device *dev) 4484 { 4485 if (dev) { 4486 #ifdef CONFIG_PCPU_DEV_REFCNT 4487 this_cpu_inc(*dev->pcpu_refcnt); 4488 #else 4489 refcount_inc(&dev->dev_refcnt); 4490 #endif 4491 } 4492 } 4493 4494 static inline void __netdev_tracker_alloc(struct net_device *dev, 4495 netdevice_tracker *tracker, 4496 gfp_t gfp) 4497 { 4498 #ifdef CONFIG_NET_DEV_REFCNT_TRACKER 4499 ref_tracker_alloc(&dev->refcnt_tracker, tracker, gfp); 4500 #endif 4501 } 4502 4503 /* netdev_tracker_alloc() can upgrade a prior untracked reference 4504 * taken by dev_get_by_name()/dev_get_by_index() to a tracked one. 4505 */ 4506 static inline void netdev_tracker_alloc(struct net_device *dev, 4507 netdevice_tracker *tracker, gfp_t gfp) 4508 { 4509 #ifdef CONFIG_NET_DEV_REFCNT_TRACKER 4510 refcount_dec(&dev->refcnt_tracker.no_tracker); 4511 __netdev_tracker_alloc(dev, tracker, gfp); 4512 #endif 4513 } 4514 4515 static inline void netdev_tracker_free(struct net_device *dev, 4516 netdevice_tracker *tracker) 4517 { 4518 #ifdef CONFIG_NET_DEV_REFCNT_TRACKER 4519 ref_tracker_free(&dev->refcnt_tracker, tracker); 4520 #endif 4521 } 4522 4523 static inline void netdev_hold(struct net_device *dev, 4524 netdevice_tracker *tracker, gfp_t gfp) 4525 { 4526 if (dev) { 4527 __dev_hold(dev); 4528 __netdev_tracker_alloc(dev, tracker, gfp); 4529 } 4530 } 4531 4532 static inline void netdev_put(struct net_device *dev, 4533 netdevice_tracker *tracker) 4534 { 4535 if (dev) { 4536 netdev_tracker_free(dev, tracker); 4537 __dev_put(dev); 4538 } 4539 } 4540 4541 /** 4542 * dev_hold - get reference to device 4543 * @dev: network device 4544 * 4545 * Hold reference to device to keep it from being freed. 4546 * Try using netdev_hold() instead. 4547 */ 4548 static inline void dev_hold(struct net_device *dev) 4549 { 4550 netdev_hold(dev, NULL, GFP_ATOMIC); 4551 } 4552 4553 /** 4554 * dev_put - release reference to device 4555 * @dev: network device 4556 * 4557 * Release reference to device to allow it to be freed. 4558 * Try using netdev_put() instead. 4559 */ 4560 static inline void dev_put(struct net_device *dev) 4561 { 4562 netdev_put(dev, NULL); 4563 } 4564 4565 DEFINE_FREE(dev_put, struct net_device *, if (_T) dev_put(_T)) 4566 4567 static inline void netdev_ref_replace(struct net_device *odev, 4568 struct net_device *ndev, 4569 netdevice_tracker *tracker, 4570 gfp_t gfp) 4571 { 4572 if (odev) 4573 netdev_tracker_free(odev, tracker); 4574 4575 __dev_hold(ndev); 4576 __dev_put(odev); 4577 4578 if (ndev) 4579 __netdev_tracker_alloc(ndev, tracker, gfp); 4580 } 4581 4582 /* Carrier loss detection, dial on demand. The functions netif_carrier_on 4583 * and _off may be called from IRQ context, but it is caller 4584 * who is responsible for serialization of these calls. 4585 * 4586 * The name carrier is inappropriate, these functions should really be 4587 * called netif_lowerlayer_*() because they represent the state of any 4588 * kind of lower layer not just hardware media. 4589 */ 4590 void linkwatch_fire_event(struct net_device *dev); 4591 4592 /** 4593 * linkwatch_sync_dev - sync linkwatch for the given device 4594 * @dev: network device to sync linkwatch for 4595 * 4596 * Sync linkwatch for the given device, removing it from the 4597 * pending work list (if queued). 4598 */ 4599 void linkwatch_sync_dev(struct net_device *dev); 4600 void __linkwatch_sync_dev(struct net_device *dev); 4601 4602 /** 4603 * netif_carrier_ok - test if carrier present 4604 * @dev: network device 4605 * 4606 * Check if carrier is present on device 4607 */ 4608 static inline bool netif_carrier_ok(const struct net_device *dev) 4609 { 4610 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state); 4611 } 4612 4613 unsigned long dev_trans_start(struct net_device *dev); 4614 4615 void netdev_watchdog_up(struct net_device *dev); 4616 4617 void netif_carrier_on(struct net_device *dev); 4618 void netif_carrier_off(struct net_device *dev); 4619 void netif_carrier_event(struct net_device *dev); 4620 4621 /** 4622 * netif_dormant_on - mark device as dormant. 4623 * @dev: network device 4624 * 4625 * Mark device as dormant (as per RFC2863). 4626 * 4627 * The dormant state indicates that the relevant interface is not 4628 * actually in a condition to pass packets (i.e., it is not 'up') but is 4629 * in a "pending" state, waiting for some external event. For "on- 4630 * demand" interfaces, this new state identifies the situation where the 4631 * interface is waiting for events to place it in the up state. 4632 */ 4633 static inline void netif_dormant_on(struct net_device *dev) 4634 { 4635 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state)) 4636 linkwatch_fire_event(dev); 4637 } 4638 4639 /** 4640 * netif_dormant_off - set device as not dormant. 4641 * @dev: network device 4642 * 4643 * Device is not in dormant state. 4644 */ 4645 static inline void netif_dormant_off(struct net_device *dev) 4646 { 4647 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state)) 4648 linkwatch_fire_event(dev); 4649 } 4650 4651 /** 4652 * netif_dormant - test if device is dormant 4653 * @dev: network device 4654 * 4655 * Check if device is dormant. 4656 */ 4657 static inline bool netif_dormant(const struct net_device *dev) 4658 { 4659 return test_bit(__LINK_STATE_DORMANT, &dev->state); 4660 } 4661 4662 4663 /** 4664 * netif_testing_on - mark device as under test. 4665 * @dev: network device 4666 * 4667 * Mark device as under test (as per RFC2863). 4668 * 4669 * The testing state indicates that some test(s) must be performed on 4670 * the interface. After completion, of the test, the interface state 4671 * will change to up, dormant, or down, as appropriate. 4672 */ 4673 static inline void netif_testing_on(struct net_device *dev) 4674 { 4675 if (!test_and_set_bit(__LINK_STATE_TESTING, &dev->state)) 4676 linkwatch_fire_event(dev); 4677 } 4678 4679 /** 4680 * netif_testing_off - set device as not under test. 4681 * @dev: network device 4682 * 4683 * Device is not in testing state. 4684 */ 4685 static inline void netif_testing_off(struct net_device *dev) 4686 { 4687 if (test_and_clear_bit(__LINK_STATE_TESTING, &dev->state)) 4688 linkwatch_fire_event(dev); 4689 } 4690 4691 /** 4692 * netif_testing - test if device is under test 4693 * @dev: network device 4694 * 4695 * Check if device is under test 4696 */ 4697 static inline bool netif_testing(const struct net_device *dev) 4698 { 4699 return test_bit(__LINK_STATE_TESTING, &dev->state); 4700 } 4701 4702 4703 /** 4704 * netif_oper_up - test if device is operational 4705 * @dev: network device 4706 * 4707 * Check if carrier is operational 4708 */ 4709 static inline bool netif_oper_up(const struct net_device *dev) 4710 { 4711 unsigned int operstate = READ_ONCE(dev->operstate); 4712 4713 return operstate == IF_OPER_UP || 4714 operstate == IF_OPER_UNKNOWN /* backward compat */; 4715 } 4716 4717 /** 4718 * netif_device_present - is device available or removed 4719 * @dev: network device 4720 * 4721 * Check if device has not been removed from system. 4722 */ 4723 static inline bool netif_device_present(const struct net_device *dev) 4724 { 4725 return test_bit(__LINK_STATE_PRESENT, &dev->state); 4726 } 4727 4728 void netif_device_detach(struct net_device *dev); 4729 4730 void netif_device_attach(struct net_device *dev); 4731 4732 /* 4733 * Network interface message level settings 4734 */ 4735 4736 enum { 4737 NETIF_MSG_DRV_BIT, 4738 NETIF_MSG_PROBE_BIT, 4739 NETIF_MSG_LINK_BIT, 4740 NETIF_MSG_TIMER_BIT, 4741 NETIF_MSG_IFDOWN_BIT, 4742 NETIF_MSG_IFUP_BIT, 4743 NETIF_MSG_RX_ERR_BIT, 4744 NETIF_MSG_TX_ERR_BIT, 4745 NETIF_MSG_TX_QUEUED_BIT, 4746 NETIF_MSG_INTR_BIT, 4747 NETIF_MSG_TX_DONE_BIT, 4748 NETIF_MSG_RX_STATUS_BIT, 4749 NETIF_MSG_PKTDATA_BIT, 4750 NETIF_MSG_HW_BIT, 4751 NETIF_MSG_WOL_BIT, 4752 4753 /* When you add a new bit above, update netif_msg_class_names array 4754 * in net/ethtool/common.c 4755 */ 4756 NETIF_MSG_CLASS_COUNT, 4757 }; 4758 /* Both ethtool_ops interface and internal driver implementation use u32 */ 4759 static_assert(NETIF_MSG_CLASS_COUNT <= 32); 4760 4761 #define __NETIF_MSG_BIT(bit) ((u32)1 << (bit)) 4762 #define __NETIF_MSG(name) __NETIF_MSG_BIT(NETIF_MSG_ ## name ## _BIT) 4763 4764 #define NETIF_MSG_DRV __NETIF_MSG(DRV) 4765 #define NETIF_MSG_PROBE __NETIF_MSG(PROBE) 4766 #define NETIF_MSG_LINK __NETIF_MSG(LINK) 4767 #define NETIF_MSG_TIMER __NETIF_MSG(TIMER) 4768 #define NETIF_MSG_IFDOWN __NETIF_MSG(IFDOWN) 4769 #define NETIF_MSG_IFUP __NETIF_MSG(IFUP) 4770 #define NETIF_MSG_RX_ERR __NETIF_MSG(RX_ERR) 4771 #define NETIF_MSG_TX_ERR __NETIF_MSG(TX_ERR) 4772 #define NETIF_MSG_TX_QUEUED __NETIF_MSG(TX_QUEUED) 4773 #define NETIF_MSG_INTR __NETIF_MSG(INTR) 4774 #define NETIF_MSG_TX_DONE __NETIF_MSG(TX_DONE) 4775 #define NETIF_MSG_RX_STATUS __NETIF_MSG(RX_STATUS) 4776 #define NETIF_MSG_PKTDATA __NETIF_MSG(PKTDATA) 4777 #define NETIF_MSG_HW __NETIF_MSG(HW) 4778 #define NETIF_MSG_WOL __NETIF_MSG(WOL) 4779 4780 #define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV) 4781 #define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE) 4782 #define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK) 4783 #define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER) 4784 #define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN) 4785 #define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP) 4786 #define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR) 4787 #define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR) 4788 #define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED) 4789 #define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR) 4790 #define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE) 4791 #define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS) 4792 #define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA) 4793 #define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW) 4794 #define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL) 4795 4796 static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits) 4797 { 4798 /* use default */ 4799 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8)) 4800 return default_msg_enable_bits; 4801 if (debug_value == 0) /* no output */ 4802 return 0; 4803 /* set low N bits */ 4804 return (1U << debug_value) - 1; 4805 } 4806 4807 static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu) 4808 { 4809 spin_lock(&txq->_xmit_lock); 4810 /* Pairs with READ_ONCE() in netif_tx_owned() */ 4811 WRITE_ONCE(txq->xmit_lock_owner, cpu); 4812 } 4813 4814 static inline bool __netif_tx_acquire(struct netdev_queue *txq) 4815 { 4816 __acquire(&txq->_xmit_lock); 4817 return true; 4818 } 4819 4820 static inline void __netif_tx_release(struct netdev_queue *txq) 4821 { 4822 __release(&txq->_xmit_lock); 4823 } 4824 4825 static inline void __netif_tx_lock_bh(struct netdev_queue *txq) 4826 { 4827 spin_lock_bh(&txq->_xmit_lock); 4828 /* Pairs with READ_ONCE() in netif_tx_owned() */ 4829 WRITE_ONCE(txq->xmit_lock_owner, smp_processor_id()); 4830 } 4831 4832 static inline bool __netif_tx_trylock(struct netdev_queue *txq) 4833 { 4834 bool ok = spin_trylock(&txq->_xmit_lock); 4835 4836 if (likely(ok)) { 4837 /* Pairs with READ_ONCE() in netif_tx_owned() */ 4838 WRITE_ONCE(txq->xmit_lock_owner, smp_processor_id()); 4839 } 4840 return ok; 4841 } 4842 4843 static inline void __netif_tx_unlock(struct netdev_queue *txq) 4844 { 4845 /* Pairs with READ_ONCE() in netif_tx_owned() */ 4846 WRITE_ONCE(txq->xmit_lock_owner, -1); 4847 spin_unlock(&txq->_xmit_lock); 4848 } 4849 4850 static inline void __netif_tx_unlock_bh(struct netdev_queue *txq) 4851 { 4852 /* Pairs with READ_ONCE() in netif_tx_owned() */ 4853 WRITE_ONCE(txq->xmit_lock_owner, -1); 4854 spin_unlock_bh(&txq->_xmit_lock); 4855 } 4856 4857 /* 4858 * txq->trans_start can be read locklessly from dev_watchdog() 4859 */ 4860 static inline void txq_trans_update(const struct net_device *dev, 4861 struct netdev_queue *txq) 4862 { 4863 if (!dev->lltx) 4864 WRITE_ONCE(txq->trans_start, jiffies); 4865 } 4866 4867 static inline void txq_trans_cond_update(struct netdev_queue *txq) 4868 { 4869 unsigned long now = jiffies; 4870 4871 if (READ_ONCE(txq->trans_start) != now) 4872 WRITE_ONCE(txq->trans_start, now); 4873 } 4874 4875 /* legacy drivers only, netdev_start_xmit() sets txq->trans_start */ 4876 static inline void netif_trans_update(struct net_device *dev) 4877 { 4878 struct netdev_queue *txq = netdev_get_tx_queue(dev, 0); 4879 4880 txq_trans_cond_update(txq); 4881 } 4882 4883 /** 4884 * netif_tx_lock - grab network device transmit lock 4885 * @dev: network device 4886 * 4887 * Get network device transmit lock 4888 */ 4889 void netif_tx_lock(struct net_device *dev); 4890 4891 static inline void netif_tx_lock_bh(struct net_device *dev) 4892 { 4893 local_bh_disable(); 4894 netif_tx_lock(dev); 4895 } 4896 4897 void netif_tx_unlock(struct net_device *dev); 4898 4899 static inline void netif_tx_unlock_bh(struct net_device *dev) 4900 { 4901 netif_tx_unlock(dev); 4902 local_bh_enable(); 4903 } 4904 4905 #define HARD_TX_LOCK(dev, txq, cpu) { \ 4906 if (!(dev)->lltx) { \ 4907 __netif_tx_lock(txq, cpu); \ 4908 } else { \ 4909 __netif_tx_acquire(txq); \ 4910 } \ 4911 } 4912 4913 #define HARD_TX_TRYLOCK(dev, txq) \ 4914 (!(dev)->lltx ? \ 4915 __netif_tx_trylock(txq) : \ 4916 __netif_tx_acquire(txq)) 4917 4918 #define HARD_TX_UNLOCK(dev, txq) { \ 4919 if (!(dev)->lltx) { \ 4920 __netif_tx_unlock(txq); \ 4921 } else { \ 4922 __netif_tx_release(txq); \ 4923 } \ 4924 } 4925 4926 static inline void netif_tx_disable(struct net_device *dev) 4927 { 4928 unsigned int i; 4929 int cpu; 4930 4931 local_bh_disable(); 4932 cpu = smp_processor_id(); 4933 spin_lock(&dev->tx_global_lock); 4934 for (i = 0; i < dev->num_tx_queues; i++) { 4935 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 4936 4937 __netif_tx_lock(txq, cpu); 4938 netif_tx_stop_queue(txq); 4939 __netif_tx_unlock(txq); 4940 } 4941 spin_unlock(&dev->tx_global_lock); 4942 local_bh_enable(); 4943 } 4944 4945 #ifndef CONFIG_PREEMPT_RT 4946 static inline bool netif_tx_owned(struct netdev_queue *txq, unsigned int cpu) 4947 { 4948 /* Other cpus might concurrently change txq->xmit_lock_owner 4949 * to -1 or to their cpu id, but not to our id. 4950 */ 4951 return READ_ONCE(txq->xmit_lock_owner) == cpu; 4952 } 4953 4954 #else 4955 static inline bool netif_tx_owned(struct netdev_queue *txq, unsigned int cpu) 4956 { 4957 return rt_mutex_owner(&txq->_xmit_lock.lock) == current; 4958 } 4959 4960 #endif 4961 4962 static inline void netif_addr_lock(struct net_device *dev) 4963 { 4964 unsigned char nest_level = 0; 4965 4966 #ifdef CONFIG_LOCKDEP 4967 nest_level = dev->nested_level; 4968 #endif 4969 spin_lock_nested(&dev->addr_list_lock, nest_level); 4970 } 4971 4972 static inline void netif_addr_lock_bh(struct net_device *dev) 4973 { 4974 unsigned char nest_level = 0; 4975 4976 #ifdef CONFIG_LOCKDEP 4977 nest_level = dev->nested_level; 4978 #endif 4979 local_bh_disable(); 4980 spin_lock_nested(&dev->addr_list_lock, nest_level); 4981 } 4982 4983 static inline void netif_addr_unlock(struct net_device *dev) 4984 { 4985 spin_unlock(&dev->addr_list_lock); 4986 } 4987 4988 static inline void netif_addr_unlock_bh(struct net_device *dev) 4989 { 4990 spin_unlock_bh(&dev->addr_list_lock); 4991 } 4992 4993 /* 4994 * dev_addrs walker. Should be used only for read access. Call with 4995 * rcu_read_lock held. 4996 */ 4997 #define for_each_dev_addr(dev, ha) \ 4998 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list) 4999 5000 /* These functions live elsewhere (drivers/net/net_init.c, but related) */ 5001 5002 void ether_setup(struct net_device *dev); 5003 5004 /* Allocate dummy net_device */ 5005 struct net_device *alloc_netdev_dummy(int sizeof_priv); 5006 5007 /* Support for loadable net-drivers */ 5008 struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name, 5009 unsigned char name_assign_type, 5010 void (*setup)(struct net_device *), 5011 unsigned int txqs, unsigned int rxqs); 5012 #define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \ 5013 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1) 5014 5015 #define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \ 5016 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \ 5017 count) 5018 5019 int register_netdev(struct net_device *dev); 5020 void unregister_netdev(struct net_device *dev); 5021 5022 int devm_register_netdev(struct device *dev, struct net_device *ndev); 5023 5024 /* General hardware address lists handling functions */ 5025 int __hw_addr_sync(struct netdev_hw_addr_list *to_list, 5026 struct netdev_hw_addr_list *from_list, int addr_len); 5027 int __hw_addr_sync_multiple(struct netdev_hw_addr_list *to_list, 5028 struct netdev_hw_addr_list *from_list, 5029 int addr_len); 5030 void __hw_addr_unsync(struct netdev_hw_addr_list *to_list, 5031 struct netdev_hw_addr_list *from_list, int addr_len); 5032 int __hw_addr_sync_dev(struct netdev_hw_addr_list *list, 5033 struct net_device *dev, 5034 int (*sync)(struct net_device *, const unsigned char *), 5035 int (*unsync)(struct net_device *, 5036 const unsigned char *)); 5037 int __hw_addr_ref_sync_dev(struct netdev_hw_addr_list *list, 5038 struct net_device *dev, 5039 int (*sync)(struct net_device *, 5040 const unsigned char *, int), 5041 int (*unsync)(struct net_device *, 5042 const unsigned char *, int)); 5043 void __hw_addr_ref_unsync_dev(struct netdev_hw_addr_list *list, 5044 struct net_device *dev, 5045 int (*unsync)(struct net_device *, 5046 const unsigned char *, int)); 5047 void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list, 5048 struct net_device *dev, 5049 int (*unsync)(struct net_device *, 5050 const unsigned char *)); 5051 void __hw_addr_init(struct netdev_hw_addr_list *list); 5052 void __hw_addr_flush(struct netdev_hw_addr_list *list); 5053 int __hw_addr_list_snapshot(struct netdev_hw_addr_list *snap, 5054 const struct netdev_hw_addr_list *list, 5055 int addr_len, struct netdev_hw_addr_list *cache); 5056 void __hw_addr_list_reconcile(struct netdev_hw_addr_list *real_list, 5057 struct netdev_hw_addr_list *work, 5058 struct netdev_hw_addr_list *ref, int addr_len, 5059 struct netdev_hw_addr_list *cache); 5060 5061 /* Functions used for device addresses handling */ 5062 void dev_addr_mod(struct net_device *dev, unsigned int offset, 5063 const void *addr, size_t len); 5064 5065 static inline void 5066 __dev_addr_set(struct net_device *dev, const void *addr, size_t len) 5067 { 5068 dev_addr_mod(dev, 0, addr, len); 5069 } 5070 5071 static inline void dev_addr_set(struct net_device *dev, const u8 *addr) 5072 { 5073 __dev_addr_set(dev, addr, dev->addr_len); 5074 } 5075 5076 int dev_addr_add(struct net_device *dev, const unsigned char *addr, 5077 unsigned char addr_type); 5078 int dev_addr_del(struct net_device *dev, const unsigned char *addr, 5079 unsigned char addr_type); 5080 5081 /* Functions used for unicast addresses handling */ 5082 int dev_uc_add(struct net_device *dev, const unsigned char *addr); 5083 int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr); 5084 int dev_uc_del(struct net_device *dev, const unsigned char *addr); 5085 int dev_uc_sync(struct net_device *to, struct net_device *from); 5086 int dev_uc_sync_multiple(struct net_device *to, struct net_device *from); 5087 void dev_uc_unsync(struct net_device *to, struct net_device *from); 5088 void dev_uc_flush(struct net_device *dev); 5089 void dev_uc_init(struct net_device *dev); 5090 5091 /** 5092 * __dev_uc_sync - Synchronize device's unicast list 5093 * @dev: device to sync 5094 * @sync: function to call if address should be added 5095 * @unsync: function to call if address should be removed 5096 * 5097 * Add newly added addresses to the interface, and release 5098 * addresses that have been deleted. 5099 */ 5100 static inline int __dev_uc_sync(struct net_device *dev, 5101 int (*sync)(struct net_device *, 5102 const unsigned char *), 5103 int (*unsync)(struct net_device *, 5104 const unsigned char *)) 5105 { 5106 return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync); 5107 } 5108 5109 /** 5110 * __dev_uc_unsync - Remove synchronized addresses from device 5111 * @dev: device to sync 5112 * @unsync: function to call if address should be removed 5113 * 5114 * Remove all addresses that were added to the device by dev_uc_sync(). 5115 */ 5116 static inline void __dev_uc_unsync(struct net_device *dev, 5117 int (*unsync)(struct net_device *, 5118 const unsigned char *)) 5119 { 5120 __hw_addr_unsync_dev(&dev->uc, dev, unsync); 5121 } 5122 5123 /* Functions used for multicast addresses handling */ 5124 int dev_mc_add(struct net_device *dev, const unsigned char *addr); 5125 int dev_mc_add_global(struct net_device *dev, const unsigned char *addr); 5126 int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr); 5127 int dev_mc_del(struct net_device *dev, const unsigned char *addr); 5128 int dev_mc_del_global(struct net_device *dev, const unsigned char *addr); 5129 int dev_mc_sync(struct net_device *to, struct net_device *from); 5130 int dev_mc_sync_multiple(struct net_device *to, struct net_device *from); 5131 void dev_mc_unsync(struct net_device *to, struct net_device *from); 5132 void dev_mc_flush(struct net_device *dev); 5133 void dev_mc_init(struct net_device *dev); 5134 5135 /** 5136 * __dev_mc_sync - Synchronize device's multicast list 5137 * @dev: device to sync 5138 * @sync: function to call if address should be added 5139 * @unsync: function to call if address should be removed 5140 * 5141 * Add newly added addresses to the interface, and release 5142 * addresses that have been deleted. 5143 */ 5144 static inline int __dev_mc_sync(struct net_device *dev, 5145 int (*sync)(struct net_device *, 5146 const unsigned char *), 5147 int (*unsync)(struct net_device *, 5148 const unsigned char *)) 5149 { 5150 return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync); 5151 } 5152 5153 /** 5154 * __dev_mc_unsync - Remove synchronized addresses from device 5155 * @dev: device to sync 5156 * @unsync: function to call if address should be removed 5157 * 5158 * Remove all addresses that were added to the device by dev_mc_sync(). 5159 */ 5160 static inline void __dev_mc_unsync(struct net_device *dev, 5161 int (*unsync)(struct net_device *, 5162 const unsigned char *)) 5163 { 5164 __hw_addr_unsync_dev(&dev->mc, dev, unsync); 5165 } 5166 5167 /* Functions used for secondary unicast and multicast support */ 5168 void dev_set_rx_mode(struct net_device *dev); 5169 void netif_rx_mode_schedule_retry(struct net_device *dev); 5170 int netif_set_promiscuity(struct net_device *dev, int inc); 5171 int dev_set_promiscuity(struct net_device *dev, int inc); 5172 int netif_set_allmulti(struct net_device *dev, int inc, bool notify); 5173 int dev_set_allmulti(struct net_device *dev, int inc); 5174 void netif_state_change(struct net_device *dev); 5175 void netdev_state_change(struct net_device *dev); 5176 void __netdev_notify_peers(struct net_device *dev); 5177 void netdev_notify_peers(struct net_device *dev); 5178 void netdev_features_change(struct net_device *dev); 5179 /* Load a device via the kmod */ 5180 void dev_load(struct net *net, const char *name); 5181 struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev, 5182 struct rtnl_link_stats64 *storage); 5183 void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64, 5184 const struct net_device_stats *netdev_stats); 5185 void dev_fetch_sw_netstats(struct rtnl_link_stats64 *s, 5186 const struct pcpu_sw_netstats __percpu *netstats); 5187 void dev_get_tstats64(struct net_device *dev, struct rtnl_link_stats64 *s); 5188 5189 void netdev_work_sched(struct net_device *dev, unsigned long events); 5190 unsigned long netdev_work_cancel(struct net_device *dev, unsigned long mask); 5191 5192 enum { 5193 NESTED_SYNC_IMM_BIT, 5194 NESTED_SYNC_TODO_BIT, 5195 }; 5196 5197 #define __NESTED_SYNC_BIT(bit) ((u32)1 << (bit)) 5198 #define __NESTED_SYNC(name) __NESTED_SYNC_BIT(NESTED_SYNC_ ## name ## _BIT) 5199 5200 #define NESTED_SYNC_IMM __NESTED_SYNC(IMM) 5201 #define NESTED_SYNC_TODO __NESTED_SYNC(TODO) 5202 5203 struct netdev_nested_priv { 5204 unsigned char flags; 5205 void *data; 5206 }; 5207 5208 bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev); 5209 struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev, 5210 struct list_head **iter); 5211 5212 /* iterate through upper list, must be called under RCU read lock */ 5213 #define netdev_for_each_upper_dev_rcu(dev, updev, iter) \ 5214 for (iter = &(dev)->adj_list.upper, \ 5215 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \ 5216 updev; \ 5217 updev = netdev_upper_get_next_dev_rcu(dev, &(iter))) 5218 5219 int netdev_walk_all_upper_dev_rcu(struct net_device *dev, 5220 int (*fn)(struct net_device *upper_dev, 5221 struct netdev_nested_priv *priv), 5222 struct netdev_nested_priv *priv); 5223 5224 bool netdev_has_upper_dev_all_rcu(struct net_device *dev, 5225 struct net_device *upper_dev); 5226 5227 bool netdev_has_any_upper_dev(struct net_device *dev); 5228 5229 void *netdev_lower_get_next_private(struct net_device *dev, 5230 struct list_head **iter); 5231 void *netdev_lower_get_next_private_rcu(struct net_device *dev, 5232 struct list_head **iter); 5233 5234 #define netdev_for_each_lower_private(dev, priv, iter) \ 5235 for (iter = (dev)->adj_list.lower.next, \ 5236 priv = netdev_lower_get_next_private(dev, &(iter)); \ 5237 priv; \ 5238 priv = netdev_lower_get_next_private(dev, &(iter))) 5239 5240 #define netdev_for_each_lower_private_rcu(dev, priv, iter) \ 5241 for (iter = &(dev)->adj_list.lower, \ 5242 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \ 5243 priv; \ 5244 priv = netdev_lower_get_next_private_rcu(dev, &(iter))) 5245 5246 void *netdev_lower_get_next(struct net_device *dev, 5247 struct list_head **iter); 5248 5249 #define netdev_for_each_lower_dev(dev, ldev, iter) \ 5250 for (iter = (dev)->adj_list.lower.next, \ 5251 ldev = netdev_lower_get_next(dev, &(iter)); \ 5252 ldev; \ 5253 ldev = netdev_lower_get_next(dev, &(iter))) 5254 5255 struct net_device *netdev_next_lower_dev_rcu(struct net_device *dev, 5256 struct list_head **iter); 5257 int netdev_walk_all_lower_dev(struct net_device *dev, 5258 int (*fn)(struct net_device *lower_dev, 5259 struct netdev_nested_priv *priv), 5260 struct netdev_nested_priv *priv); 5261 int netdev_walk_all_lower_dev_rcu(struct net_device *dev, 5262 int (*fn)(struct net_device *lower_dev, 5263 struct netdev_nested_priv *priv), 5264 struct netdev_nested_priv *priv); 5265 5266 void *netdev_adjacent_get_private(struct list_head *adj_list); 5267 void *netdev_lower_get_first_private_rcu(struct net_device *dev); 5268 struct net_device *netdev_master_upper_dev_get(struct net_device *dev); 5269 struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev); 5270 int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev, 5271 struct netlink_ext_ack *extack); 5272 int netdev_master_upper_dev_link(struct net_device *dev, 5273 struct net_device *upper_dev, 5274 void *upper_priv, void *upper_info, 5275 struct netlink_ext_ack *extack); 5276 void netdev_upper_dev_unlink(struct net_device *dev, 5277 struct net_device *upper_dev); 5278 int netdev_adjacent_change_prepare(struct net_device *old_dev, 5279 struct net_device *new_dev, 5280 struct net_device *dev, 5281 struct netlink_ext_ack *extack); 5282 void netdev_adjacent_change_commit(struct net_device *old_dev, 5283 struct net_device *new_dev, 5284 struct net_device *dev); 5285 void netdev_adjacent_change_abort(struct net_device *old_dev, 5286 struct net_device *new_dev, 5287 struct net_device *dev); 5288 void netdev_adjacent_rename_links(struct net_device *dev, char *oldname); 5289 void *netdev_lower_dev_get_private(struct net_device *dev, 5290 struct net_device *lower_dev); 5291 void netdev_lower_state_changed(struct net_device *lower_dev, 5292 void *lower_state_info); 5293 5294 #define NETDEV_RSS_KEY_LEN 256 5295 extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN] __read_mostly; 5296 void netdev_rss_key_fill(void *buffer, size_t len); 5297 5298 int skb_checksum_help(struct sk_buff *skb); 5299 int skb_crc32c_csum_help(struct sk_buff *skb); 5300 int skb_csum_hwoffload_help(struct sk_buff *skb, 5301 const netdev_features_t features); 5302 5303 struct netdev_bonding_info { 5304 ifslave slave; 5305 ifbond master; 5306 }; 5307 5308 struct netdev_notifier_bonding_info { 5309 struct netdev_notifier_info info; /* must be first */ 5310 struct netdev_bonding_info bonding_info; 5311 }; 5312 5313 void netdev_bonding_info_change(struct net_device *dev, 5314 struct netdev_bonding_info *bonding_info); 5315 5316 #if IS_ENABLED(CONFIG_ETHTOOL_NETLINK) 5317 void ethtool_notify(struct net_device *dev, unsigned int cmd); 5318 #else 5319 static inline void ethtool_notify(struct net_device *dev, unsigned int cmd) 5320 { 5321 } 5322 #endif 5323 5324 __be16 skb_network_protocol(struct sk_buff *skb, int *depth); 5325 5326 static inline bool can_checksum_protocol(netdev_features_t features, 5327 __be16 protocol) 5328 { 5329 if (protocol == htons(ETH_P_FCOE)) 5330 return !!(features & NETIF_F_FCOE_CRC); 5331 5332 /* Assume this is an IP checksum (not SCTP CRC) */ 5333 5334 if (features & NETIF_F_HW_CSUM) { 5335 /* Can checksum everything */ 5336 return true; 5337 } 5338 5339 switch (protocol) { 5340 case htons(ETH_P_IP): 5341 return !!(features & NETIF_F_IP_CSUM); 5342 case htons(ETH_P_IPV6): 5343 return !!(features & NETIF_F_IPV6_CSUM); 5344 default: 5345 return false; 5346 } 5347 } 5348 5349 #ifdef CONFIG_BUG 5350 void netdev_rx_csum_fault(struct net_device *dev, struct sk_buff *skb); 5351 #else 5352 static inline void netdev_rx_csum_fault(struct net_device *dev, 5353 struct sk_buff *skb) 5354 { 5355 } 5356 #endif 5357 /* rx skb timestamps */ 5358 void net_enable_timestamp(void); 5359 void net_disable_timestamp(void); 5360 5361 static inline ktime_t netdev_get_tstamp(struct net_device *dev, 5362 const struct skb_shared_hwtstamps *hwtstamps, 5363 bool cycles) 5364 { 5365 const struct net_device_ops *ops = dev->netdev_ops; 5366 5367 if (ops->ndo_get_tstamp) 5368 return ops->ndo_get_tstamp(dev, hwtstamps, cycles); 5369 5370 return hwtstamps->hwtstamp; 5371 } 5372 5373 #ifndef CONFIG_PREEMPT_RT 5374 static inline void netdev_xmit_set_more(bool more) 5375 { 5376 __this_cpu_write(softnet_data.xmit.more, more); 5377 } 5378 5379 static inline bool netdev_xmit_more(void) 5380 { 5381 return __this_cpu_read(softnet_data.xmit.more); 5382 } 5383 #else 5384 static inline void netdev_xmit_set_more(bool more) 5385 { 5386 current->net_xmit.more = more; 5387 } 5388 5389 static inline bool netdev_xmit_more(void) 5390 { 5391 return current->net_xmit.more; 5392 } 5393 #endif 5394 5395 static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops, 5396 struct sk_buff *skb, struct net_device *dev, 5397 bool more) 5398 { 5399 netdev_xmit_set_more(more); 5400 return ops->ndo_start_xmit(skb, dev); 5401 } 5402 5403 static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev, 5404 struct netdev_queue *txq, bool more) 5405 { 5406 const struct net_device_ops *ops = dev->netdev_ops; 5407 netdev_tx_t rc; 5408 5409 rc = __netdev_start_xmit(ops, skb, dev, more); 5410 if (rc == NETDEV_TX_OK) 5411 txq_trans_update(dev, txq); 5412 5413 return rc; 5414 } 5415 5416 int netdev_class_create_file_ns(const struct class_attribute *class_attr, 5417 const struct ns_common *ns); 5418 void netdev_class_remove_file_ns(const struct class_attribute *class_attr, 5419 const struct ns_common *ns); 5420 5421 extern const struct kobj_ns_type_operations net_ns_type_operations; 5422 5423 const char *netdev_drivername(const struct net_device *dev); 5424 5425 static inline netdev_features_t netdev_intersect_features(netdev_features_t f1, 5426 netdev_features_t f2) 5427 { 5428 if ((f1 ^ f2) & NETIF_F_HW_CSUM) { 5429 if (f1 & NETIF_F_HW_CSUM) 5430 f1 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM); 5431 else 5432 f2 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM); 5433 } 5434 5435 return f1 & f2; 5436 } 5437 5438 static inline netdev_features_t netdev_get_wanted_features( 5439 struct net_device *dev) 5440 { 5441 return (dev->features & ~dev->hw_features) | dev->wanted_features; 5442 } 5443 netdev_features_t netdev_increment_features(netdev_features_t all, 5444 netdev_features_t one, netdev_features_t mask); 5445 5446 /* Allow TSO being used on stacked device : 5447 * Performing the GSO segmentation before last device 5448 * is a performance improvement. 5449 */ 5450 static inline netdev_features_t netdev_add_tso_features(netdev_features_t features, 5451 netdev_features_t mask) 5452 { 5453 return netdev_increment_features(features, NETIF_F_ALL_TSO | 5454 NETIF_F_ALL_FOR_ALL, mask); 5455 } 5456 5457 int __netdev_update_features(struct net_device *dev); 5458 void netdev_update_features(struct net_device *dev); 5459 void netdev_change_features(struct net_device *dev); 5460 void netdev_compute_master_upper_features(struct net_device *dev, bool update_header); 5461 5462 void netif_stacked_transfer_operstate(const struct net_device *rootdev, 5463 struct net_device *dev); 5464 5465 netdev_features_t passthru_features_check(struct sk_buff *skb, 5466 struct net_device *dev, 5467 netdev_features_t features); 5468 netdev_features_t netif_skb_features(struct sk_buff *skb); 5469 void skb_warn_bad_offload(const struct sk_buff *skb); 5470 5471 static inline bool net_gso_ok(netdev_features_t features, int gso_type) 5472 { 5473 netdev_features_t feature; 5474 5475 if (gso_type & (SKB_GSO_TCP_FIXEDID | SKB_GSO_TCP_FIXEDID_INNER)) 5476 gso_type |= __SKB_GSO_TCP_FIXEDID; 5477 5478 feature = ((netdev_features_t)gso_type << NETIF_F_GSO_SHIFT) & NETIF_F_GSO_MASK; 5479 5480 /* check flags correspondence */ 5481 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT)); 5482 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT)); 5483 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT)); 5484 BUILD_BUG_ON(__SKB_GSO_TCP_FIXEDID != (NETIF_F_TSO_MANGLEID >> NETIF_F_GSO_SHIFT)); 5485 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT)); 5486 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT)); 5487 BUILD_BUG_ON(SKB_GSO_GRE != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT)); 5488 BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT)); 5489 BUILD_BUG_ON(SKB_GSO_IPXIP4 != (NETIF_F_GSO_IPXIP4 >> NETIF_F_GSO_SHIFT)); 5490 BUILD_BUG_ON(SKB_GSO_IPXIP6 != (NETIF_F_GSO_IPXIP6 >> NETIF_F_GSO_SHIFT)); 5491 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT)); 5492 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT)); 5493 BUILD_BUG_ON(SKB_GSO_PARTIAL != (NETIF_F_GSO_PARTIAL >> NETIF_F_GSO_SHIFT)); 5494 BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT)); 5495 BUILD_BUG_ON(SKB_GSO_SCTP != (NETIF_F_GSO_SCTP >> NETIF_F_GSO_SHIFT)); 5496 BUILD_BUG_ON(SKB_GSO_ESP != (NETIF_F_GSO_ESP >> NETIF_F_GSO_SHIFT)); 5497 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_GSO_UDP >> NETIF_F_GSO_SHIFT)); 5498 BUILD_BUG_ON(SKB_GSO_UDP_L4 != (NETIF_F_GSO_UDP_L4 >> NETIF_F_GSO_SHIFT)); 5499 BUILD_BUG_ON(SKB_GSO_FRAGLIST != (NETIF_F_GSO_FRAGLIST >> NETIF_F_GSO_SHIFT)); 5500 BUILD_BUG_ON(SKB_GSO_TCP_ACCECN != 5501 (NETIF_F_GSO_ACCECN >> NETIF_F_GSO_SHIFT)); 5502 5503 return (features & feature) == feature; 5504 } 5505 5506 static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features) 5507 { 5508 return net_gso_ok(features, skb_shinfo(skb)->gso_type) && 5509 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST)); 5510 } 5511 5512 static inline bool netif_needs_gso(struct sk_buff *skb, 5513 netdev_features_t features) 5514 { 5515 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) || 5516 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) && 5517 (skb->ip_summed != CHECKSUM_UNNECESSARY))); 5518 } 5519 5520 void netif_set_tso_max_size(struct net_device *dev, unsigned int size); 5521 void netif_set_tso_max_segs(struct net_device *dev, unsigned int segs); 5522 void netif_inherit_tso_max(struct net_device *to, 5523 const struct net_device *from); 5524 5525 static inline unsigned int 5526 netif_get_gro_max_size(const struct net_device *dev, const struct sk_buff *skb) 5527 { 5528 /* pairs with WRITE_ONCE() in netif_set_gro(_ipv4)_max_size() */ 5529 return skb->protocol == htons(ETH_P_IPV6) ? 5530 READ_ONCE(dev->gro_max_size) : 5531 READ_ONCE(dev->gro_ipv4_max_size); 5532 } 5533 5534 static inline unsigned int 5535 netif_get_gso_max_size(const struct net_device *dev, const struct sk_buff *skb) 5536 { 5537 /* pairs with WRITE_ONCE() in netif_set_gso(_ipv4)_max_size() */ 5538 return skb->protocol == htons(ETH_P_IPV6) ? 5539 READ_ONCE(dev->gso_max_size) : 5540 READ_ONCE(dev->gso_ipv4_max_size); 5541 } 5542 5543 static inline bool netif_is_macsec(const struct net_device *dev) 5544 { 5545 return dev->priv_flags & IFF_MACSEC; 5546 } 5547 5548 static inline bool netif_is_macvlan(const struct net_device *dev) 5549 { 5550 return dev->priv_flags & IFF_MACVLAN; 5551 } 5552 5553 static inline bool netif_is_macvlan_port(const struct net_device *dev) 5554 { 5555 return dev->priv_flags & IFF_MACVLAN_PORT; 5556 } 5557 5558 static inline bool netif_is_bond_master(const struct net_device *dev) 5559 { 5560 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING; 5561 } 5562 5563 static inline bool netif_is_bond_slave(const struct net_device *dev) 5564 { 5565 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING; 5566 } 5567 5568 static inline bool netif_supports_nofcs(struct net_device *dev) 5569 { 5570 return dev->priv_flags & IFF_SUPP_NOFCS; 5571 } 5572 5573 static inline bool netif_has_l3_rx_handler(const struct net_device *dev) 5574 { 5575 return dev->priv_flags & IFF_L3MDEV_RX_HANDLER; 5576 } 5577 5578 static inline bool netif_is_l3_master(const struct net_device *dev) 5579 { 5580 return dev->priv_flags & IFF_L3MDEV_MASTER; 5581 } 5582 5583 static inline bool netif_is_l3_slave(const struct net_device *dev) 5584 { 5585 return dev->priv_flags & IFF_L3MDEV_SLAVE; 5586 } 5587 5588 static inline int dev_sdif(const struct net_device *dev) 5589 { 5590 #ifdef CONFIG_NET_L3_MASTER_DEV 5591 if (netif_is_l3_slave(dev)) 5592 return dev->ifindex; 5593 #endif 5594 return 0; 5595 } 5596 5597 static inline bool netif_is_bridge_master(const struct net_device *dev) 5598 { 5599 return dev->priv_flags & IFF_EBRIDGE; 5600 } 5601 5602 static inline bool netif_is_bridge_port(const struct net_device *dev) 5603 { 5604 return dev->priv_flags & IFF_BRIDGE_PORT; 5605 } 5606 5607 static inline bool netif_is_ovs_master(const struct net_device *dev) 5608 { 5609 return dev->priv_flags & IFF_OPENVSWITCH; 5610 } 5611 5612 static inline bool netif_is_ovs_port(const struct net_device *dev) 5613 { 5614 return dev->priv_flags & IFF_OVS_DATAPATH; 5615 } 5616 5617 static inline bool netif_is_any_bridge_master(const struct net_device *dev) 5618 { 5619 return netif_is_bridge_master(dev) || netif_is_ovs_master(dev); 5620 } 5621 5622 static inline bool netif_is_any_bridge_port(const struct net_device *dev) 5623 { 5624 return netif_is_bridge_port(dev) || netif_is_ovs_port(dev); 5625 } 5626 5627 static inline bool netif_is_team_master(const struct net_device *dev) 5628 { 5629 return dev->priv_flags & IFF_TEAM; 5630 } 5631 5632 static inline bool netif_is_team_port(const struct net_device *dev) 5633 { 5634 return dev->priv_flags & IFF_TEAM_PORT; 5635 } 5636 5637 static inline bool netif_is_lag_master(const struct net_device *dev) 5638 { 5639 return netif_is_bond_master(dev) || netif_is_team_master(dev); 5640 } 5641 5642 static inline bool netif_is_lag_port(const struct net_device *dev) 5643 { 5644 return netif_is_bond_slave(dev) || netif_is_team_port(dev); 5645 } 5646 5647 bool netif_is_rxfh_configured(const struct net_device *dev); 5648 5649 static inline bool netif_is_failover(const struct net_device *dev) 5650 { 5651 return dev->priv_flags & IFF_FAILOVER; 5652 } 5653 5654 static inline bool netif_is_failover_slave(const struct net_device *dev) 5655 { 5656 return dev->priv_flags & IFF_FAILOVER_SLAVE; 5657 } 5658 5659 /* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */ 5660 static inline void netif_keep_dst(struct net_device *dev) 5661 { 5662 dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM); 5663 } 5664 5665 /* return true if dev can't cope with mtu frames that need vlan tag insertion */ 5666 static inline bool netif_reduces_vlan_mtu(struct net_device *dev) 5667 { 5668 /* TODO: reserve and use an additional IFF bit, if we get more users */ 5669 return netif_is_macsec(dev); 5670 } 5671 5672 extern struct pernet_operations __net_initdata loopback_net_ops; 5673 5674 /* Logging, debugging and troubleshooting/diagnostic helpers. */ 5675 5676 /* netdev_printk helpers, similar to dev_printk */ 5677 5678 static inline const char *netdev_name(const struct net_device *dev) 5679 { 5680 if (!dev->name[0] || strchr(dev->name, '%')) 5681 return "(unnamed net_device)"; 5682 return dev->name; 5683 } 5684 5685 static inline const char *netdev_reg_state(const struct net_device *dev) 5686 { 5687 u8 reg_state = READ_ONCE(dev->reg_state); 5688 5689 switch (reg_state) { 5690 case NETREG_UNINITIALIZED: return " (uninitialized)"; 5691 case NETREG_REGISTERED: return ""; 5692 case NETREG_UNREGISTERING: return " (unregistering)"; 5693 case NETREG_UNREGISTERED: return " (unregistered)"; 5694 case NETREG_RELEASED: return " (released)"; 5695 case NETREG_DUMMY: return " (dummy)"; 5696 } 5697 5698 WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, reg_state); 5699 return " (unknown)"; 5700 } 5701 5702 #define MODULE_ALIAS_NETDEV(device) \ 5703 MODULE_ALIAS("netdev-" device) 5704 5705 /* 5706 * netdev_WARN() acts like dev_printk(), but with the key difference 5707 * of using a WARN/WARN_ON to get the message out, including the 5708 * file/line information and a backtrace. 5709 */ 5710 #define netdev_WARN(dev, format, args...) \ 5711 WARN(1, "netdevice: %s%s: " format, netdev_name(dev), \ 5712 netdev_reg_state(dev), ##args) 5713 5714 #define netdev_WARN_ONCE(dev, format, args...) \ 5715 WARN_ONCE(1, "netdevice: %s%s: " format, netdev_name(dev), \ 5716 netdev_reg_state(dev), ##args) 5717 5718 /* 5719 * The list of packet types we will receive (as opposed to discard) 5720 * and the routines to invoke. 5721 * 5722 * Why 16. Because with 16 the only overlap we get on a hash of the 5723 * low nibble of the protocol value is RARP/SNAP/X.25. 5724 * 5725 * 0800 IP 5726 * 0001 802.3 5727 * 0002 AX.25 5728 * 0004 802.2 5729 * 8035 RARP 5730 * 0005 SNAP 5731 * 0805 X.25 5732 * 0806 ARP 5733 * 8137 IPX 5734 * 0009 Localtalk 5735 * 86DD IPv6 5736 */ 5737 #define PTYPE_HASH_SIZE (16) 5738 #define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1) 5739 5740 extern struct list_head ptype_base[PTYPE_HASH_SIZE] __read_mostly; 5741 5742 extern struct net_device *blackhole_netdev; 5743 5744 /* Note: Avoid these macros in fast path, prefer per-cpu or per-queue counters. */ 5745 #define DEV_STATS_INC(DEV, FIELD) atomic_long_inc(&(DEV)->stats.__##FIELD) 5746 #define DEV_STATS_ADD(DEV, FIELD, VAL) \ 5747 atomic_long_add((VAL), &(DEV)->stats.__##FIELD) 5748 #define DEV_STATS_READ(DEV, FIELD) atomic_long_read(&(DEV)->stats.__##FIELD) 5749 5750 #endif /* _LINUX_NETDEVICE_H */ 5751