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