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, <[email protected]> 13 * Corey Minyard <[email protected]> 14 * Donald J. Becker, <[email protected]> 15 * Alan Cox, <[email protected]> 16 * Bjorn Ekwall. <[email protected]> 17 * Pekka Riikonen <[email protected]> 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 32 #include <linux/percpu.h> 33 #include <linux/rculist.h> 34 #include <linux/workqueue.h> 35 #include <linux/dynamic_queue_limits.h> 36 37 #include <net/net_namespace.h> 38 #ifdef CONFIG_DCB 39 #include <net/dcbnl.h> 40 #endif 41 #include <net/netprio_cgroup.h> 42 #include <net/xdp.h> 43 44 #include <linux/netdev_features.h> 45 #include <linux/neighbour.h> 46 #include <uapi/linux/netdevice.h> 47 #include <uapi/linux/if_bonding.h> 48 #include <uapi/linux/pkt_cls.h> 49 #include <linux/hashtable.h> 50 #include <linux/rbtree.h> 51 #include <linux/ref_tracker.h> 52 53 struct netpoll_info; 54 struct device; 55 struct ethtool_ops; 56 struct phy_device; 57 struct dsa_port; 58 struct ip_tunnel_parm; 59 struct macsec_context; 60 struct macsec_ops; 61 62 struct sfp_bus; 63 /* 802.11 specific */ 64 struct wireless_dev; 65 /* 802.15.4 specific */ 66 struct wpan_dev; 67 struct mpls_dev; 68 /* UDP Tunnel offloads */ 69 struct udp_tunnel_info; 70 struct udp_tunnel_nic_info; 71 struct udp_tunnel_nic; 72 struct bpf_prog; 73 struct xdp_buff; 74 75 void synchronize_net(void); 76 void netdev_set_default_ethtool_ops(struct net_device *dev, 77 const struct ethtool_ops *ops); 78 79 /* Backlog congestion levels */ 80 #define NET_RX_SUCCESS 0 /* keep 'em coming, baby */ 81 #define NET_RX_DROP 1 /* packet dropped */ 82 83 #define MAX_NEST_DEV 8 84 85 /* 86 * Transmit return codes: transmit return codes originate from three different 87 * namespaces: 88 * 89 * - qdisc return codes 90 * - driver transmit return codes 91 * - errno values 92 * 93 * Drivers are allowed to return any one of those in their hard_start_xmit() 94 * function. Real network devices commonly used with qdiscs should only return 95 * the driver transmit return codes though - when qdiscs are used, the actual 96 * transmission happens asynchronously, so the value is not propagated to 97 * higher layers. Virtual network devices transmit synchronously; in this case 98 * the driver transmit return codes are consumed by dev_queue_xmit(), and all 99 * others are propagated to higher layers. 100 */ 101 102 /* qdisc ->enqueue() return codes. */ 103 #define NET_XMIT_SUCCESS 0x00 104 #define NET_XMIT_DROP 0x01 /* skb dropped */ 105 #define NET_XMIT_CN 0x02 /* congestion notification */ 106 #define NET_XMIT_MASK 0x0f /* qdisc flags in net/sch_generic.h */ 107 108 /* NET_XMIT_CN is special. It does not guarantee that this packet is lost. It 109 * indicates that the device will soon be dropping packets, or already drops 110 * some packets of the same priority; prompting us to send less aggressively. */ 111 #define net_xmit_eval(e) ((e) == NET_XMIT_CN ? 0 : (e)) 112 #define net_xmit_errno(e) ((e) != NET_XMIT_CN ? -ENOBUFS : 0) 113 114 /* Driver transmit return codes */ 115 #define NETDEV_TX_MASK 0xf0 116 117 enum netdev_tx { 118 __NETDEV_TX_MIN = INT_MIN, /* make sure enum is signed */ 119 NETDEV_TX_OK = 0x00, /* driver took care of packet */ 120 NETDEV_TX_BUSY = 0x10, /* driver tx path was busy*/ 121 }; 122 typedef enum netdev_tx netdev_tx_t; 123 124 /* 125 * Current order: NETDEV_TX_MASK > NET_XMIT_MASK >= 0 is significant; 126 * hard_start_xmit() return < NET_XMIT_MASK means skb was consumed. 127 */ 128 static inline bool dev_xmit_complete(int rc) 129 { 130 /* 131 * Positive cases with an skb consumed by a driver: 132 * - successful transmission (rc == NETDEV_TX_OK) 133 * - error while transmitting (rc < 0) 134 * - error while queueing to a different device (rc & NET_XMIT_MASK) 135 */ 136 if (likely(rc < NET_XMIT_MASK)) 137 return true; 138 139 return false; 140 } 141 142 /* 143 * Compute the worst-case header length according to the protocols 144 * used. 145 */ 146 147 #if defined(CONFIG_HYPERV_NET) 148 # define LL_MAX_HEADER 128 149 #elif defined(CONFIG_WLAN) || IS_ENABLED(CONFIG_AX25) 150 # if defined(CONFIG_MAC80211_MESH) 151 # define LL_MAX_HEADER 128 152 # else 153 # define LL_MAX_HEADER 96 154 # endif 155 #else 156 # define LL_MAX_HEADER 32 157 #endif 158 159 #if !IS_ENABLED(CONFIG_NET_IPIP) && !IS_ENABLED(CONFIG_NET_IPGRE) && \ 160 !IS_ENABLED(CONFIG_IPV6_SIT) && !IS_ENABLED(CONFIG_IPV6_TUNNEL) 161 #define MAX_HEADER LL_MAX_HEADER 162 #else 163 #define MAX_HEADER (LL_MAX_HEADER + 48) 164 #endif 165 166 /* 167 * Old network device statistics. Fields are native words 168 * (unsigned long) so they can be read and written atomically. 169 */ 170 171 struct net_device_stats { 172 unsigned long rx_packets; 173 unsigned long tx_packets; 174 unsigned long rx_bytes; 175 unsigned long tx_bytes; 176 unsigned long rx_errors; 177 unsigned long tx_errors; 178 unsigned long rx_dropped; 179 unsigned long tx_dropped; 180 unsigned long multicast; 181 unsigned long collisions; 182 unsigned long rx_length_errors; 183 unsigned long rx_over_errors; 184 unsigned long rx_crc_errors; 185 unsigned long rx_frame_errors; 186 unsigned long rx_fifo_errors; 187 unsigned long rx_missed_errors; 188 unsigned long tx_aborted_errors; 189 unsigned long tx_carrier_errors; 190 unsigned long tx_fifo_errors; 191 unsigned long tx_heartbeat_errors; 192 unsigned long tx_window_errors; 193 unsigned long rx_compressed; 194 unsigned long tx_compressed; 195 }; 196 197 198 #include <linux/cache.h> 199 #include <linux/skbuff.h> 200 201 #ifdef CONFIG_RPS 202 #include <linux/static_key.h> 203 extern struct static_key_false rps_needed; 204 extern struct static_key_false rfs_needed; 205 #endif 206 207 struct neighbour; 208 struct neigh_parms; 209 struct sk_buff; 210 211 struct netdev_hw_addr { 212 struct list_head list; 213 struct rb_node node; 214 unsigned char addr[MAX_ADDR_LEN]; 215 unsigned char type; 216 #define NETDEV_HW_ADDR_T_LAN 1 217 #define NETDEV_HW_ADDR_T_SAN 2 218 #define NETDEV_HW_ADDR_T_UNICAST 3 219 #define NETDEV_HW_ADDR_T_MULTICAST 4 220 bool global_use; 221 int sync_cnt; 222 int refcount; 223 int synced; 224 struct rcu_head rcu_head; 225 }; 226 227 struct netdev_hw_addr_list { 228 struct list_head list; 229 int count; 230 231 /* Auxiliary tree for faster lookup on addition and deletion */ 232 struct rb_root tree; 233 }; 234 235 #define netdev_hw_addr_list_count(l) ((l)->count) 236 #define netdev_hw_addr_list_empty(l) (netdev_hw_addr_list_count(l) == 0) 237 #define netdev_hw_addr_list_for_each(ha, l) \ 238 list_for_each_entry(ha, &(l)->list, list) 239 240 #define netdev_uc_count(dev) netdev_hw_addr_list_count(&(dev)->uc) 241 #define netdev_uc_empty(dev) netdev_hw_addr_list_empty(&(dev)->uc) 242 #define netdev_for_each_uc_addr(ha, dev) \ 243 netdev_hw_addr_list_for_each(ha, &(dev)->uc) 244 245 #define netdev_mc_count(dev) netdev_hw_addr_list_count(&(dev)->mc) 246 #define netdev_mc_empty(dev) netdev_hw_addr_list_empty(&(dev)->mc) 247 #define netdev_for_each_mc_addr(ha, dev) \ 248 netdev_hw_addr_list_for_each(ha, &(dev)->mc) 249 250 struct hh_cache { 251 unsigned int hh_len; 252 seqlock_t hh_lock; 253 254 /* cached hardware header; allow for machine alignment needs. */ 255 #define HH_DATA_MOD 16 256 #define HH_DATA_OFF(__len) \ 257 (HH_DATA_MOD - (((__len - 1) & (HH_DATA_MOD - 1)) + 1)) 258 #define HH_DATA_ALIGN(__len) \ 259 (((__len)+(HH_DATA_MOD-1))&~(HH_DATA_MOD - 1)) 260 unsigned long hh_data[HH_DATA_ALIGN(LL_MAX_HEADER) / sizeof(long)]; 261 }; 262 263 /* Reserve HH_DATA_MOD byte-aligned hard_header_len, but at least that much. 264 * Alternative is: 265 * dev->hard_header_len ? (dev->hard_header_len + 266 * (HH_DATA_MOD - 1)) & ~(HH_DATA_MOD - 1) : 0 267 * 268 * We could use other alignment values, but we must maintain the 269 * relationship HH alignment <= LL alignment. 270 */ 271 #define LL_RESERVED_SPACE(dev) \ 272 ((((dev)->hard_header_len+(dev)->needed_headroom)&~(HH_DATA_MOD - 1)) + HH_DATA_MOD) 273 #define LL_RESERVED_SPACE_EXTRA(dev,extra) \ 274 ((((dev)->hard_header_len+(dev)->needed_headroom+(extra))&~(HH_DATA_MOD - 1)) + HH_DATA_MOD) 275 276 struct header_ops { 277 int (*create) (struct sk_buff *skb, struct net_device *dev, 278 unsigned short type, const void *daddr, 279 const void *saddr, unsigned int len); 280 int (*parse)(const struct sk_buff *skb, unsigned char *haddr); 281 int (*cache)(const struct neighbour *neigh, struct hh_cache *hh, __be16 type); 282 void (*cache_update)(struct hh_cache *hh, 283 const struct net_device *dev, 284 const unsigned char *haddr); 285 bool (*validate)(const char *ll_header, unsigned int len); 286 __be16 (*parse_protocol)(const struct sk_buff *skb); 287 }; 288 289 /* These flag bits are private to the generic network queueing 290 * layer; they may not be explicitly referenced by any other 291 * code. 292 */ 293 294 enum netdev_state_t { 295 __LINK_STATE_START, 296 __LINK_STATE_PRESENT, 297 __LINK_STATE_NOCARRIER, 298 __LINK_STATE_LINKWATCH_PENDING, 299 __LINK_STATE_DORMANT, 300 __LINK_STATE_TESTING, 301 }; 302 303 304 #ifdef CONFIG_NET_DEV_REFCNT_TRACKER 305 typedef struct ref_tracker *netdevice_tracker; 306 #else 307 typedef struct {} netdevice_tracker; 308 #endif 309 310 struct gro_list { 311 struct list_head list; 312 int count; 313 }; 314 315 /* 316 * size of gro hash buckets, must less than bit number of 317 * napi_struct::gro_bitmask 318 */ 319 #define GRO_HASH_BUCKETS 8 320 321 /* 322 * Structure for NAPI scheduling similar to tasklet but with weighting 323 */ 324 struct napi_struct { 325 /* The poll_list must only be managed by the entity which 326 * changes the state of the NAPI_STATE_SCHED bit. This means 327 * whoever atomically sets that bit can add this napi_struct 328 * to the per-CPU poll_list, and whoever clears that bit 329 * can remove from the list right before clearing the bit. 330 */ 331 struct list_head poll_list; 332 333 unsigned long state; 334 int weight; 335 int defer_hard_irqs_count; 336 unsigned long gro_bitmask; 337 int (*poll)(struct napi_struct *, int); 338 #ifdef CONFIG_NETPOLL 339 int poll_owner; 340 #endif 341 struct net_device *dev; 342 struct gro_list gro_hash[GRO_HASH_BUCKETS]; 343 struct sk_buff *skb; 344 struct list_head rx_list; /* Pending GRO_NORMAL skbs */ 345 int rx_count; /* length of rx_list */ 346 struct hrtimer timer; 347 struct list_head dev_list; 348 struct hlist_node napi_hash_node; 349 unsigned int napi_id; 350 struct task_struct *thread; 351 }; 352 353 enum { 354 NAPI_STATE_SCHED, /* Poll is scheduled */ 355 NAPI_STATE_MISSED, /* reschedule a napi */ 356 NAPI_STATE_DISABLE, /* Disable pending */ 357 NAPI_STATE_NPSVC, /* Netpoll - don't dequeue from poll_list */ 358 NAPI_STATE_LISTED, /* NAPI added to system lists */ 359 NAPI_STATE_NO_BUSY_POLL, /* Do not add in napi_hash, no busy polling */ 360 NAPI_STATE_IN_BUSY_POLL, /* sk_busy_loop() owns this NAPI */ 361 NAPI_STATE_PREFER_BUSY_POLL, /* prefer busy-polling over softirq processing*/ 362 NAPI_STATE_THREADED, /* The poll is performed inside its own thread*/ 363 NAPI_STATE_SCHED_THREADED, /* Napi is currently scheduled in threaded mode */ 364 }; 365 366 enum { 367 NAPIF_STATE_SCHED = BIT(NAPI_STATE_SCHED), 368 NAPIF_STATE_MISSED = BIT(NAPI_STATE_MISSED), 369 NAPIF_STATE_DISABLE = BIT(NAPI_STATE_DISABLE), 370 NAPIF_STATE_NPSVC = BIT(NAPI_STATE_NPSVC), 371 NAPIF_STATE_LISTED = BIT(NAPI_STATE_LISTED), 372 NAPIF_STATE_NO_BUSY_POLL = BIT(NAPI_STATE_NO_BUSY_POLL), 373 NAPIF_STATE_IN_BUSY_POLL = BIT(NAPI_STATE_IN_BUSY_POLL), 374 NAPIF_STATE_PREFER_BUSY_POLL = BIT(NAPI_STATE_PREFER_BUSY_POLL), 375 NAPIF_STATE_THREADED = BIT(NAPI_STATE_THREADED), 376 NAPIF_STATE_SCHED_THREADED = BIT(NAPI_STATE_SCHED_THREADED), 377 }; 378 379 enum gro_result { 380 GRO_MERGED, 381 GRO_MERGED_FREE, 382 GRO_HELD, 383 GRO_NORMAL, 384 GRO_CONSUMED, 385 }; 386 typedef enum gro_result gro_result_t; 387 388 /* 389 * enum rx_handler_result - Possible return values for rx_handlers. 390 * @RX_HANDLER_CONSUMED: skb was consumed by rx_handler, do not process it 391 * further. 392 * @RX_HANDLER_ANOTHER: Do another round in receive path. This is indicated in 393 * case skb->dev was changed by rx_handler. 394 * @RX_HANDLER_EXACT: Force exact delivery, no wildcard. 395 * @RX_HANDLER_PASS: Do nothing, pass the skb as if no rx_handler was called. 396 * 397 * rx_handlers are functions called from inside __netif_receive_skb(), to do 398 * special processing of the skb, prior to delivery to protocol handlers. 399 * 400 * Currently, a net_device can only have a single rx_handler registered. Trying 401 * to register a second rx_handler will return -EBUSY. 402 * 403 * To register a rx_handler on a net_device, use netdev_rx_handler_register(). 404 * To unregister a rx_handler on a net_device, use 405 * netdev_rx_handler_unregister(). 406 * 407 * Upon return, rx_handler is expected to tell __netif_receive_skb() what to 408 * do with the skb. 409 * 410 * If the rx_handler consumed the skb in some way, it should return 411 * RX_HANDLER_CONSUMED. This is appropriate when the rx_handler arranged for 412 * the skb to be delivered in some other way. 413 * 414 * If the rx_handler changed skb->dev, to divert the skb to another 415 * net_device, it should return RX_HANDLER_ANOTHER. The rx_handler for the 416 * new device will be called if it exists. 417 * 418 * If the rx_handler decides the skb should be ignored, it should return 419 * RX_HANDLER_EXACT. The skb will only be delivered to protocol handlers that 420 * are registered on exact device (ptype->dev == skb->dev). 421 * 422 * If the rx_handler didn't change skb->dev, but wants the skb to be normally 423 * delivered, it should return RX_HANDLER_PASS. 424 * 425 * A device without a registered rx_handler will behave as if rx_handler 426 * returned RX_HANDLER_PASS. 427 */ 428 429 enum rx_handler_result { 430 RX_HANDLER_CONSUMED, 431 RX_HANDLER_ANOTHER, 432 RX_HANDLER_EXACT, 433 RX_HANDLER_PASS, 434 }; 435 typedef enum rx_handler_result rx_handler_result_t; 436 typedef rx_handler_result_t rx_handler_func_t(struct sk_buff **pskb); 437 438 void __napi_schedule(struct napi_struct *n); 439 void __napi_schedule_irqoff(struct napi_struct *n); 440 441 static inline bool napi_disable_pending(struct napi_struct *n) 442 { 443 return test_bit(NAPI_STATE_DISABLE, &n->state); 444 } 445 446 static inline bool napi_prefer_busy_poll(struct napi_struct *n) 447 { 448 return test_bit(NAPI_STATE_PREFER_BUSY_POLL, &n->state); 449 } 450 451 bool napi_schedule_prep(struct napi_struct *n); 452 453 /** 454 * napi_schedule - schedule NAPI poll 455 * @n: NAPI context 456 * 457 * Schedule NAPI poll routine to be called if it is not already 458 * running. 459 */ 460 static inline void napi_schedule(struct napi_struct *n) 461 { 462 if (napi_schedule_prep(n)) 463 __napi_schedule(n); 464 } 465 466 /** 467 * napi_schedule_irqoff - schedule NAPI poll 468 * @n: NAPI context 469 * 470 * Variant of napi_schedule(), assuming hard irqs are masked. 471 */ 472 static inline void napi_schedule_irqoff(struct napi_struct *n) 473 { 474 if (napi_schedule_prep(n)) 475 __napi_schedule_irqoff(n); 476 } 477 478 /* Try to reschedule poll. Called by dev->poll() after napi_complete(). */ 479 static inline bool napi_reschedule(struct napi_struct *napi) 480 { 481 if (napi_schedule_prep(napi)) { 482 __napi_schedule(napi); 483 return true; 484 } 485 return false; 486 } 487 488 bool napi_complete_done(struct napi_struct *n, int work_done); 489 /** 490 * napi_complete - NAPI processing complete 491 * @n: NAPI context 492 * 493 * Mark NAPI processing as complete. 494 * Consider using napi_complete_done() instead. 495 * Return false if device should avoid rearming interrupts. 496 */ 497 static inline bool napi_complete(struct napi_struct *n) 498 { 499 return napi_complete_done(n, 0); 500 } 501 502 int dev_set_threaded(struct net_device *dev, bool threaded); 503 504 /** 505 * napi_disable - prevent NAPI from scheduling 506 * @n: NAPI context 507 * 508 * Stop NAPI from being scheduled on this context. 509 * Waits till any outstanding processing completes. 510 */ 511 void napi_disable(struct napi_struct *n); 512 513 void napi_enable(struct napi_struct *n); 514 515 /** 516 * napi_synchronize - wait until NAPI is not running 517 * @n: NAPI context 518 * 519 * Wait until NAPI is done being scheduled on this context. 520 * Waits till any outstanding processing completes but 521 * does not disable future activations. 522 */ 523 static inline void napi_synchronize(const struct napi_struct *n) 524 { 525 if (IS_ENABLED(CONFIG_SMP)) 526 while (test_bit(NAPI_STATE_SCHED, &n->state)) 527 msleep(1); 528 else 529 barrier(); 530 } 531 532 /** 533 * napi_if_scheduled_mark_missed - if napi is running, set the 534 * NAPIF_STATE_MISSED 535 * @n: NAPI context 536 * 537 * If napi is running, set the NAPIF_STATE_MISSED, and return true if 538 * NAPI is scheduled. 539 **/ 540 static inline bool napi_if_scheduled_mark_missed(struct napi_struct *n) 541 { 542 unsigned long val, new; 543 544 do { 545 val = READ_ONCE(n->state); 546 if (val & NAPIF_STATE_DISABLE) 547 return true; 548 549 if (!(val & NAPIF_STATE_SCHED)) 550 return false; 551 552 new = val | NAPIF_STATE_MISSED; 553 } while (cmpxchg(&n->state, val, new) != val); 554 555 return true; 556 } 557 558 enum netdev_queue_state_t { 559 __QUEUE_STATE_DRV_XOFF, 560 __QUEUE_STATE_STACK_XOFF, 561 __QUEUE_STATE_FROZEN, 562 }; 563 564 #define QUEUE_STATE_DRV_XOFF (1 << __QUEUE_STATE_DRV_XOFF) 565 #define QUEUE_STATE_STACK_XOFF (1 << __QUEUE_STATE_STACK_XOFF) 566 #define QUEUE_STATE_FROZEN (1 << __QUEUE_STATE_FROZEN) 567 568 #define QUEUE_STATE_ANY_XOFF (QUEUE_STATE_DRV_XOFF | QUEUE_STATE_STACK_XOFF) 569 #define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \ 570 QUEUE_STATE_FROZEN) 571 #define QUEUE_STATE_DRV_XOFF_OR_FROZEN (QUEUE_STATE_DRV_XOFF | \ 572 QUEUE_STATE_FROZEN) 573 574 /* 575 * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue. The 576 * netif_tx_* functions below are used to manipulate this flag. The 577 * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit 578 * queue independently. The netif_xmit_*stopped functions below are called 579 * to check if the queue has been stopped by the driver or stack (either 580 * of the XOFF bits are set in the state). Drivers should not need to call 581 * netif_xmit*stopped functions, they should only be using netif_tx_*. 582 */ 583 584 struct netdev_queue { 585 /* 586 * read-mostly part 587 */ 588 struct net_device *dev; 589 netdevice_tracker dev_tracker; 590 591 struct Qdisc __rcu *qdisc; 592 struct Qdisc *qdisc_sleeping; 593 #ifdef CONFIG_SYSFS 594 struct kobject kobj; 595 #endif 596 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA) 597 int numa_node; 598 #endif 599 unsigned long tx_maxrate; 600 /* 601 * Number of TX timeouts for this queue 602 * (/sys/class/net/DEV/Q/trans_timeout) 603 */ 604 atomic_long_t trans_timeout; 605 606 /* Subordinate device that the queue has been assigned to */ 607 struct net_device *sb_dev; 608 #ifdef CONFIG_XDP_SOCKETS 609 struct xsk_buff_pool *pool; 610 #endif 611 /* 612 * write-mostly part 613 */ 614 spinlock_t _xmit_lock ____cacheline_aligned_in_smp; 615 int xmit_lock_owner; 616 /* 617 * Time (in jiffies) of last Tx 618 */ 619 unsigned long trans_start; 620 621 unsigned long state; 622 623 #ifdef CONFIG_BQL 624 struct dql dql; 625 #endif 626 } ____cacheline_aligned_in_smp; 627 628 extern int sysctl_fb_tunnels_only_for_init_net; 629 extern int sysctl_devconf_inherit_init_net; 630 631 /* 632 * sysctl_fb_tunnels_only_for_init_net == 0 : For all netns 633 * == 1 : For initns only 634 * == 2 : For none. 635 */ 636 static inline bool net_has_fallback_tunnels(const struct net *net) 637 { 638 return !IS_ENABLED(CONFIG_SYSCTL) || 639 !sysctl_fb_tunnels_only_for_init_net || 640 (net == &init_net && sysctl_fb_tunnels_only_for_init_net == 1); 641 } 642 643 static inline int netdev_queue_numa_node_read(const struct netdev_queue *q) 644 { 645 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA) 646 return q->numa_node; 647 #else 648 return NUMA_NO_NODE; 649 #endif 650 } 651 652 static inline void netdev_queue_numa_node_write(struct netdev_queue *q, int node) 653 { 654 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA) 655 q->numa_node = node; 656 #endif 657 } 658 659 #ifdef CONFIG_RPS 660 /* 661 * This structure holds an RPS map which can be of variable length. The 662 * map is an array of CPUs. 663 */ 664 struct rps_map { 665 unsigned int len; 666 struct rcu_head rcu; 667 u16 cpus[]; 668 }; 669 #define RPS_MAP_SIZE(_num) (sizeof(struct rps_map) + ((_num) * sizeof(u16))) 670 671 /* 672 * The rps_dev_flow structure contains the mapping of a flow to a CPU, the 673 * tail pointer for that CPU's input queue at the time of last enqueue, and 674 * a hardware filter index. 675 */ 676 struct rps_dev_flow { 677 u16 cpu; 678 u16 filter; 679 unsigned int last_qtail; 680 }; 681 #define RPS_NO_FILTER 0xffff 682 683 /* 684 * The rps_dev_flow_table structure contains a table of flow mappings. 685 */ 686 struct rps_dev_flow_table { 687 unsigned int mask; 688 struct rcu_head rcu; 689 struct rps_dev_flow flows[]; 690 }; 691 #define RPS_DEV_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_dev_flow_table) + \ 692 ((_num) * sizeof(struct rps_dev_flow))) 693 694 /* 695 * The rps_sock_flow_table contains mappings of flows to the last CPU 696 * on which they were processed by the application (set in recvmsg). 697 * Each entry is a 32bit value. Upper part is the high-order bits 698 * of flow hash, lower part is CPU number. 699 * rps_cpu_mask is used to partition the space, depending on number of 700 * possible CPUs : rps_cpu_mask = roundup_pow_of_two(nr_cpu_ids) - 1 701 * For example, if 64 CPUs are possible, rps_cpu_mask = 0x3f, 702 * meaning we use 32-6=26 bits for the hash. 703 */ 704 struct rps_sock_flow_table { 705 u32 mask; 706 707 u32 ents[] ____cacheline_aligned_in_smp; 708 }; 709 #define RPS_SOCK_FLOW_TABLE_SIZE(_num) (offsetof(struct rps_sock_flow_table, ents[_num])) 710 711 #define RPS_NO_CPU 0xffff 712 713 extern u32 rps_cpu_mask; 714 extern struct rps_sock_flow_table __rcu *rps_sock_flow_table; 715 716 static inline void rps_record_sock_flow(struct rps_sock_flow_table *table, 717 u32 hash) 718 { 719 if (table && hash) { 720 unsigned int index = hash & table->mask; 721 u32 val = hash & ~rps_cpu_mask; 722 723 /* We only give a hint, preemption can change CPU under us */ 724 val |= raw_smp_processor_id(); 725 726 if (table->ents[index] != val) 727 table->ents[index] = val; 728 } 729 } 730 731 #ifdef CONFIG_RFS_ACCEL 732 bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index, u32 flow_id, 733 u16 filter_id); 734 #endif 735 #endif /* CONFIG_RPS */ 736 737 /* This structure contains an instance of an RX queue. */ 738 struct netdev_rx_queue { 739 struct xdp_rxq_info xdp_rxq; 740 #ifdef CONFIG_RPS 741 struct rps_map __rcu *rps_map; 742 struct rps_dev_flow_table __rcu *rps_flow_table; 743 #endif 744 struct kobject kobj; 745 struct net_device *dev; 746 netdevice_tracker dev_tracker; 747 748 #ifdef CONFIG_XDP_SOCKETS 749 struct xsk_buff_pool *pool; 750 #endif 751 } ____cacheline_aligned_in_smp; 752 753 /* 754 * RX queue sysfs structures and functions. 755 */ 756 struct rx_queue_attribute { 757 struct attribute attr; 758 ssize_t (*show)(struct netdev_rx_queue *queue, char *buf); 759 ssize_t (*store)(struct netdev_rx_queue *queue, 760 const char *buf, size_t len); 761 }; 762 763 /* XPS map type and offset of the xps map within net_device->xps_maps[]. */ 764 enum xps_map_type { 765 XPS_CPUS = 0, 766 XPS_RXQS, 767 XPS_MAPS_MAX, 768 }; 769 770 #ifdef CONFIG_XPS 771 /* 772 * This structure holds an XPS map which can be of variable length. The 773 * map is an array of queues. 774 */ 775 struct xps_map { 776 unsigned int len; 777 unsigned int alloc_len; 778 struct rcu_head rcu; 779 u16 queues[]; 780 }; 781 #define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16))) 782 #define XPS_MIN_MAP_ALLOC ((L1_CACHE_ALIGN(offsetof(struct xps_map, queues[1])) \ 783 - sizeof(struct xps_map)) / sizeof(u16)) 784 785 /* 786 * This structure holds all XPS maps for device. Maps are indexed by CPU. 787 * 788 * We keep track of the number of cpus/rxqs used when the struct is allocated, 789 * in nr_ids. This will help not accessing out-of-bound memory. 790 * 791 * We keep track of the number of traffic classes used when the struct is 792 * allocated, in num_tc. This will be used to navigate the maps, to ensure we're 793 * not crossing its upper bound, as the original dev->num_tc can be updated in 794 * the meantime. 795 */ 796 struct xps_dev_maps { 797 struct rcu_head rcu; 798 unsigned int nr_ids; 799 s16 num_tc; 800 struct xps_map __rcu *attr_map[]; /* Either CPUs map or RXQs map */ 801 }; 802 803 #define XPS_CPU_DEV_MAPS_SIZE(_tcs) (sizeof(struct xps_dev_maps) + \ 804 (nr_cpu_ids * (_tcs) * sizeof(struct xps_map *))) 805 806 #define XPS_RXQ_DEV_MAPS_SIZE(_tcs, _rxqs) (sizeof(struct xps_dev_maps) +\ 807 (_rxqs * (_tcs) * sizeof(struct xps_map *))) 808 809 #endif /* CONFIG_XPS */ 810 811 #define TC_MAX_QUEUE 16 812 #define TC_BITMASK 15 813 /* HW offloaded queuing disciplines txq count and offset maps */ 814 struct netdev_tc_txq { 815 u16 count; 816 u16 offset; 817 }; 818 819 #if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE) 820 /* 821 * This structure is to hold information about the device 822 * configured to run FCoE protocol stack. 823 */ 824 struct netdev_fcoe_hbainfo { 825 char manufacturer[64]; 826 char serial_number[64]; 827 char hardware_version[64]; 828 char driver_version[64]; 829 char optionrom_version[64]; 830 char firmware_version[64]; 831 char model[256]; 832 char model_description[256]; 833 }; 834 #endif 835 836 #define MAX_PHYS_ITEM_ID_LEN 32 837 838 /* This structure holds a unique identifier to identify some 839 * physical item (port for example) used by a netdevice. 840 */ 841 struct netdev_phys_item_id { 842 unsigned char id[MAX_PHYS_ITEM_ID_LEN]; 843 unsigned char id_len; 844 }; 845 846 static inline bool netdev_phys_item_id_same(struct netdev_phys_item_id *a, 847 struct netdev_phys_item_id *b) 848 { 849 return a->id_len == b->id_len && 850 memcmp(a->id, b->id, a->id_len) == 0; 851 } 852 853 typedef u16 (*select_queue_fallback_t)(struct net_device *dev, 854 struct sk_buff *skb, 855 struct net_device *sb_dev); 856 857 enum net_device_path_type { 858 DEV_PATH_ETHERNET = 0, 859 DEV_PATH_VLAN, 860 DEV_PATH_BRIDGE, 861 DEV_PATH_PPPOE, 862 DEV_PATH_DSA, 863 }; 864 865 struct net_device_path { 866 enum net_device_path_type type; 867 const struct net_device *dev; 868 union { 869 struct { 870 u16 id; 871 __be16 proto; 872 u8 h_dest[ETH_ALEN]; 873 } encap; 874 struct { 875 enum { 876 DEV_PATH_BR_VLAN_KEEP, 877 DEV_PATH_BR_VLAN_TAG, 878 DEV_PATH_BR_VLAN_UNTAG, 879 DEV_PATH_BR_VLAN_UNTAG_HW, 880 } vlan_mode; 881 u16 vlan_id; 882 __be16 vlan_proto; 883 } bridge; 884 struct { 885 int port; 886 u16 proto; 887 } dsa; 888 }; 889 }; 890 891 #define NET_DEVICE_PATH_STACK_MAX 5 892 #define NET_DEVICE_PATH_VLAN_MAX 2 893 894 struct net_device_path_stack { 895 int num_paths; 896 struct net_device_path path[NET_DEVICE_PATH_STACK_MAX]; 897 }; 898 899 struct net_device_path_ctx { 900 const struct net_device *dev; 901 const u8 *daddr; 902 903 int num_vlans; 904 struct { 905 u16 id; 906 __be16 proto; 907 } vlan[NET_DEVICE_PATH_VLAN_MAX]; 908 }; 909 910 enum tc_setup_type { 911 TC_SETUP_QDISC_MQPRIO, 912 TC_SETUP_CLSU32, 913 TC_SETUP_CLSFLOWER, 914 TC_SETUP_CLSMATCHALL, 915 TC_SETUP_CLSBPF, 916 TC_SETUP_BLOCK, 917 TC_SETUP_QDISC_CBS, 918 TC_SETUP_QDISC_RED, 919 TC_SETUP_QDISC_PRIO, 920 TC_SETUP_QDISC_MQ, 921 TC_SETUP_QDISC_ETF, 922 TC_SETUP_ROOT_QDISC, 923 TC_SETUP_QDISC_GRED, 924 TC_SETUP_QDISC_TAPRIO, 925 TC_SETUP_FT, 926 TC_SETUP_QDISC_ETS, 927 TC_SETUP_QDISC_TBF, 928 TC_SETUP_QDISC_FIFO, 929 TC_SETUP_QDISC_HTB, 930 }; 931 932 /* These structures hold the attributes of bpf state that are being passed 933 * to the netdevice through the bpf op. 934 */ 935 enum bpf_netdev_command { 936 /* Set or clear a bpf program used in the earliest stages of packet 937 * rx. The prog will have been loaded as BPF_PROG_TYPE_XDP. The callee 938 * is responsible for calling bpf_prog_put on any old progs that are 939 * stored. In case of error, the callee need not release the new prog 940 * reference, but on success it takes ownership and must bpf_prog_put 941 * when it is no longer used. 942 */ 943 XDP_SETUP_PROG, 944 XDP_SETUP_PROG_HW, 945 /* BPF program for offload callbacks, invoked at program load time. */ 946 BPF_OFFLOAD_MAP_ALLOC, 947 BPF_OFFLOAD_MAP_FREE, 948 XDP_SETUP_XSK_POOL, 949 }; 950 951 struct bpf_prog_offload_ops; 952 struct netlink_ext_ack; 953 struct xdp_umem; 954 struct xdp_dev_bulk_queue; 955 struct bpf_xdp_link; 956 957 enum bpf_xdp_mode { 958 XDP_MODE_SKB = 0, 959 XDP_MODE_DRV = 1, 960 XDP_MODE_HW = 2, 961 __MAX_XDP_MODE 962 }; 963 964 struct bpf_xdp_entity { 965 struct bpf_prog *prog; 966 struct bpf_xdp_link *link; 967 }; 968 969 struct netdev_bpf { 970 enum bpf_netdev_command command; 971 union { 972 /* XDP_SETUP_PROG */ 973 struct { 974 u32 flags; 975 struct bpf_prog *prog; 976 struct netlink_ext_ack *extack; 977 }; 978 /* BPF_OFFLOAD_MAP_ALLOC, BPF_OFFLOAD_MAP_FREE */ 979 struct { 980 struct bpf_offloaded_map *offmap; 981 }; 982 /* XDP_SETUP_XSK_POOL */ 983 struct { 984 struct xsk_buff_pool *pool; 985 u16 queue_id; 986 } xsk; 987 }; 988 }; 989 990 /* Flags for ndo_xsk_wakeup. */ 991 #define XDP_WAKEUP_RX (1 << 0) 992 #define XDP_WAKEUP_TX (1 << 1) 993 994 #ifdef CONFIG_XFRM_OFFLOAD 995 struct xfrmdev_ops { 996 int (*xdo_dev_state_add) (struct xfrm_state *x); 997 void (*xdo_dev_state_delete) (struct xfrm_state *x); 998 void (*xdo_dev_state_free) (struct xfrm_state *x); 999 bool (*xdo_dev_offload_ok) (struct sk_buff *skb, 1000 struct xfrm_state *x); 1001 void (*xdo_dev_state_advance_esn) (struct xfrm_state *x); 1002 }; 1003 #endif 1004 1005 struct dev_ifalias { 1006 struct rcu_head rcuhead; 1007 char ifalias[]; 1008 }; 1009 1010 struct devlink; 1011 struct tlsdev_ops; 1012 1013 struct netdev_name_node { 1014 struct hlist_node hlist; 1015 struct list_head list; 1016 struct net_device *dev; 1017 const char *name; 1018 }; 1019 1020 int netdev_name_node_alt_create(struct net_device *dev, const char *name); 1021 int netdev_name_node_alt_destroy(struct net_device *dev, const char *name); 1022 1023 struct netdev_net_notifier { 1024 struct list_head list; 1025 struct notifier_block *nb; 1026 }; 1027 1028 /* 1029 * This structure defines the management hooks for network devices. 1030 * The following hooks can be defined; unless noted otherwise, they are 1031 * optional and can be filled with a null pointer. 1032 * 1033 * int (*ndo_init)(struct net_device *dev); 1034 * This function is called once when a network device is registered. 1035 * The network device can use this for any late stage initialization 1036 * or semantic validation. It can fail with an error code which will 1037 * be propagated back to register_netdev. 1038 * 1039 * void (*ndo_uninit)(struct net_device *dev); 1040 * This function is called when device is unregistered or when registration 1041 * fails. It is not called if init fails. 1042 * 1043 * int (*ndo_open)(struct net_device *dev); 1044 * This function is called when a network device transitions to the up 1045 * state. 1046 * 1047 * int (*ndo_stop)(struct net_device *dev); 1048 * This function is called when a network device transitions to the down 1049 * state. 1050 * 1051 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb, 1052 * struct net_device *dev); 1053 * Called when a packet needs to be transmitted. 1054 * Returns NETDEV_TX_OK. Can return NETDEV_TX_BUSY, but you should stop 1055 * the queue before that can happen; it's for obsolete devices and weird 1056 * corner cases, but the stack really does a non-trivial amount 1057 * of useless work if you return NETDEV_TX_BUSY. 1058 * Required; cannot be NULL. 1059 * 1060 * netdev_features_t (*ndo_features_check)(struct sk_buff *skb, 1061 * struct net_device *dev 1062 * netdev_features_t features); 1063 * Called by core transmit path to determine if device is capable of 1064 * performing offload operations on a given packet. This is to give 1065 * the device an opportunity to implement any restrictions that cannot 1066 * be otherwise expressed by feature flags. The check is called with 1067 * the set of features that the stack has calculated and it returns 1068 * those the driver believes to be appropriate. 1069 * 1070 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb, 1071 * struct net_device *sb_dev); 1072 * Called to decide which queue to use when device supports multiple 1073 * transmit queues. 1074 * 1075 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags); 1076 * This function is called to allow device receiver to make 1077 * changes to configuration when multicast or promiscuous is enabled. 1078 * 1079 * void (*ndo_set_rx_mode)(struct net_device *dev); 1080 * This function is called device changes address list filtering. 1081 * If driver handles unicast address filtering, it should set 1082 * IFF_UNICAST_FLT in its priv_flags. 1083 * 1084 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr); 1085 * This function is called when the Media Access Control address 1086 * needs to be changed. If this interface is not defined, the 1087 * MAC address can not be changed. 1088 * 1089 * int (*ndo_validate_addr)(struct net_device *dev); 1090 * Test if Media Access Control address is valid for the device. 1091 * 1092 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd); 1093 * Old-style ioctl entry point. This is used internally by the 1094 * appletalk and ieee802154 subsystems but is no longer called by 1095 * the device ioctl handler. 1096 * 1097 * int (*ndo_siocbond)(struct net_device *dev, struct ifreq *ifr, int cmd); 1098 * Used by the bonding driver for its device specific ioctls: 1099 * SIOCBONDENSLAVE, SIOCBONDRELEASE, SIOCBONDSETHWADDR, SIOCBONDCHANGEACTIVE, 1100 * SIOCBONDSLAVEINFOQUERY, and SIOCBONDINFOQUERY 1101 * 1102 * * int (*ndo_eth_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd); 1103 * Called for ethernet specific ioctls: SIOCGMIIPHY, SIOCGMIIREG, 1104 * SIOCSMIIREG, SIOCSHWTSTAMP and SIOCGHWTSTAMP. 1105 * 1106 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map); 1107 * Used to set network devices bus interface parameters. This interface 1108 * is retained for legacy reasons; new devices should use the bus 1109 * interface (PCI) for low level management. 1110 * 1111 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu); 1112 * Called when a user wants to change the Maximum Transfer Unit 1113 * of a device. 1114 * 1115 * void (*ndo_tx_timeout)(struct net_device *dev, unsigned int txqueue); 1116 * Callback used when the transmitter has not made any progress 1117 * for dev->watchdog ticks. 1118 * 1119 * void (*ndo_get_stats64)(struct net_device *dev, 1120 * struct rtnl_link_stats64 *storage); 1121 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev); 1122 * Called when a user wants to get the network device usage 1123 * statistics. Drivers must do one of the following: 1124 * 1. Define @ndo_get_stats64 to fill in a zero-initialised 1125 * rtnl_link_stats64 structure passed by the caller. 1126 * 2. Define @ndo_get_stats to update a net_device_stats structure 1127 * (which should normally be dev->stats) and return a pointer to 1128 * it. The structure may be changed asynchronously only if each 1129 * field is written atomically. 1130 * 3. Update dev->stats asynchronously and atomically, and define 1131 * neither operation. 1132 * 1133 * bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id) 1134 * Return true if this device supports offload stats of this attr_id. 1135 * 1136 * int (*ndo_get_offload_stats)(int attr_id, const struct net_device *dev, 1137 * void *attr_data) 1138 * Get statistics for offload operations by attr_id. Write it into the 1139 * attr_data pointer. 1140 * 1141 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16 vid); 1142 * If device supports VLAN filtering this function is called when a 1143 * VLAN id is registered. 1144 * 1145 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, __be16 proto, u16 vid); 1146 * If device supports VLAN filtering this function is called when a 1147 * VLAN id is unregistered. 1148 * 1149 * void (*ndo_poll_controller)(struct net_device *dev); 1150 * 1151 * SR-IOV management functions. 1152 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac); 1153 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan, 1154 * u8 qos, __be16 proto); 1155 * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate, 1156 * int max_tx_rate); 1157 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting); 1158 * int (*ndo_set_vf_trust)(struct net_device *dev, int vf, bool setting); 1159 * int (*ndo_get_vf_config)(struct net_device *dev, 1160 * int vf, struct ifla_vf_info *ivf); 1161 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state); 1162 * int (*ndo_set_vf_port)(struct net_device *dev, int vf, 1163 * struct nlattr *port[]); 1164 * 1165 * Enable or disable the VF ability to query its RSS Redirection Table and 1166 * Hash Key. This is needed since on some devices VF share this information 1167 * with PF and querying it may introduce a theoretical security risk. 1168 * int (*ndo_set_vf_rss_query_en)(struct net_device *dev, int vf, bool setting); 1169 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb); 1170 * int (*ndo_setup_tc)(struct net_device *dev, enum tc_setup_type type, 1171 * void *type_data); 1172 * Called to setup any 'tc' scheduler, classifier or action on @dev. 1173 * This is always called from the stack with the rtnl lock held and netif 1174 * tx queues stopped. This allows the netdevice to perform queue 1175 * management safely. 1176 * 1177 * Fiber Channel over Ethernet (FCoE) offload functions. 1178 * int (*ndo_fcoe_enable)(struct net_device *dev); 1179 * Called when the FCoE protocol stack wants to start using LLD for FCoE 1180 * so the underlying device can perform whatever needed configuration or 1181 * initialization to support acceleration of FCoE traffic. 1182 * 1183 * int (*ndo_fcoe_disable)(struct net_device *dev); 1184 * Called when the FCoE protocol stack wants to stop using LLD for FCoE 1185 * so the underlying device can perform whatever needed clean-ups to 1186 * stop supporting acceleration of FCoE traffic. 1187 * 1188 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid, 1189 * struct scatterlist *sgl, unsigned int sgc); 1190 * Called when the FCoE Initiator wants to initialize an I/O that 1191 * is a possible candidate for Direct Data Placement (DDP). The LLD can 1192 * perform necessary setup and returns 1 to indicate the device is set up 1193 * successfully to perform DDP on this I/O, otherwise this returns 0. 1194 * 1195 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid); 1196 * Called when the FCoE Initiator/Target is done with the DDPed I/O as 1197 * indicated by the FC exchange id 'xid', so the underlying device can 1198 * clean up and reuse resources for later DDP requests. 1199 * 1200 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid, 1201 * struct scatterlist *sgl, unsigned int sgc); 1202 * Called when the FCoE Target wants to initialize an I/O that 1203 * is a possible candidate for Direct Data Placement (DDP). The LLD can 1204 * perform necessary setup and returns 1 to indicate the device is set up 1205 * successfully to perform DDP on this I/O, otherwise this returns 0. 1206 * 1207 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev, 1208 * struct netdev_fcoe_hbainfo *hbainfo); 1209 * Called when the FCoE Protocol stack wants information on the underlying 1210 * device. This information is utilized by the FCoE protocol stack to 1211 * register attributes with Fiber Channel management service as per the 1212 * FC-GS Fabric Device Management Information(FDMI) specification. 1213 * 1214 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type); 1215 * Called when the underlying device wants to override default World Wide 1216 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own 1217 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE 1218 * protocol stack to use. 1219 * 1220 * RFS acceleration. 1221 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb, 1222 * u16 rxq_index, u32 flow_id); 1223 * Set hardware filter for RFS. rxq_index is the target queue index; 1224 * flow_id is a flow ID to be passed to rps_may_expire_flow() later. 1225 * Return the filter ID on success, or a negative error code. 1226 * 1227 * Slave management functions (for bridge, bonding, etc). 1228 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev); 1229 * Called to make another netdev an underling. 1230 * 1231 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev); 1232 * Called to release previously enslaved netdev. 1233 * 1234 * struct net_device *(*ndo_get_xmit_slave)(struct net_device *dev, 1235 * struct sk_buff *skb, 1236 * bool all_slaves); 1237 * Get the xmit slave of master device. If all_slaves is true, function 1238 * assume all the slaves can transmit. 1239 * 1240 * Feature/offload setting functions. 1241 * netdev_features_t (*ndo_fix_features)(struct net_device *dev, 1242 * netdev_features_t features); 1243 * Adjusts the requested feature flags according to device-specific 1244 * constraints, and returns the resulting flags. Must not modify 1245 * the device state. 1246 * 1247 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features); 1248 * Called to update device configuration to new features. Passed 1249 * feature set might be less than what was returned by ndo_fix_features()). 1250 * Must return >0 or -errno if it changed dev->features itself. 1251 * 1252 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[], 1253 * struct net_device *dev, 1254 * const unsigned char *addr, u16 vid, u16 flags, 1255 * struct netlink_ext_ack *extack); 1256 * Adds an FDB entry to dev for addr. 1257 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[], 1258 * struct net_device *dev, 1259 * const unsigned char *addr, u16 vid) 1260 * Deletes the FDB entry from dev coresponding to addr. 1261 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb, 1262 * struct net_device *dev, struct net_device *filter_dev, 1263 * int *idx) 1264 * Used to add FDB entries to dump requests. Implementers should add 1265 * entries to skb and update idx with the number of entries. 1266 * 1267 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh, 1268 * u16 flags, struct netlink_ext_ack *extack) 1269 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq, 1270 * struct net_device *dev, u32 filter_mask, 1271 * int nlflags) 1272 * int (*ndo_bridge_dellink)(struct net_device *dev, struct nlmsghdr *nlh, 1273 * u16 flags); 1274 * 1275 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier); 1276 * Called to change device carrier. Soft-devices (like dummy, team, etc) 1277 * which do not represent real hardware may define this to allow their 1278 * userspace components to manage their virtual carrier state. Devices 1279 * that determine carrier state from physical hardware properties (eg 1280 * network cables) or protocol-dependent mechanisms (eg 1281 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function. 1282 * 1283 * int (*ndo_get_phys_port_id)(struct net_device *dev, 1284 * struct netdev_phys_item_id *ppid); 1285 * Called to get ID of physical port of this device. If driver does 1286 * not implement this, it is assumed that the hw is not able to have 1287 * multiple net devices on single physical port. 1288 * 1289 * int (*ndo_get_port_parent_id)(struct net_device *dev, 1290 * struct netdev_phys_item_id *ppid) 1291 * Called to get the parent ID of the physical port of this device. 1292 * 1293 * void* (*ndo_dfwd_add_station)(struct net_device *pdev, 1294 * struct net_device *dev) 1295 * Called by upper layer devices to accelerate switching or other 1296 * station functionality into hardware. 'pdev is the lowerdev 1297 * to use for the offload and 'dev' is the net device that will 1298 * back the offload. Returns a pointer to the private structure 1299 * the upper layer will maintain. 1300 * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv) 1301 * Called by upper layer device to delete the station created 1302 * by 'ndo_dfwd_add_station'. 'pdev' is the net device backing 1303 * the station and priv is the structure returned by the add 1304 * operation. 1305 * int (*ndo_set_tx_maxrate)(struct net_device *dev, 1306 * int queue_index, u32 maxrate); 1307 * Called when a user wants to set a max-rate limitation of specific 1308 * TX queue. 1309 * int (*ndo_get_iflink)(const struct net_device *dev); 1310 * Called to get the iflink value of this device. 1311 * int (*ndo_fill_metadata_dst)(struct net_device *dev, struct sk_buff *skb); 1312 * This function is used to get egress tunnel information for given skb. 1313 * This is useful for retrieving outer tunnel header parameters while 1314 * sampling packet. 1315 * void (*ndo_set_rx_headroom)(struct net_device *dev, int needed_headroom); 1316 * This function is used to specify the headroom that the skb must 1317 * consider when allocation skb during packet reception. Setting 1318 * appropriate rx headroom value allows avoiding skb head copy on 1319 * forward. Setting a negative value resets the rx headroom to the 1320 * default value. 1321 * int (*ndo_bpf)(struct net_device *dev, struct netdev_bpf *bpf); 1322 * This function is used to set or query state related to XDP on the 1323 * netdevice and manage BPF offload. See definition of 1324 * enum bpf_netdev_command for details. 1325 * int (*ndo_xdp_xmit)(struct net_device *dev, int n, struct xdp_frame **xdp, 1326 * u32 flags); 1327 * This function is used to submit @n XDP packets for transmit on a 1328 * netdevice. Returns number of frames successfully transmitted, frames 1329 * that got dropped are freed/returned via xdp_return_frame(). 1330 * Returns negative number, means general error invoking ndo, meaning 1331 * no frames were xmit'ed and core-caller will free all frames. 1332 * struct net_device *(*ndo_xdp_get_xmit_slave)(struct net_device *dev, 1333 * struct xdp_buff *xdp); 1334 * Get the xmit slave of master device based on the xdp_buff. 1335 * int (*ndo_xsk_wakeup)(struct net_device *dev, u32 queue_id, u32 flags); 1336 * This function is used to wake up the softirq, ksoftirqd or kthread 1337 * responsible for sending and/or receiving packets on a specific 1338 * queue id bound to an AF_XDP socket. The flags field specifies if 1339 * only RX, only Tx, or both should be woken up using the flags 1340 * XDP_WAKEUP_RX and XDP_WAKEUP_TX. 1341 * struct devlink_port *(*ndo_get_devlink_port)(struct net_device *dev); 1342 * Get devlink port instance associated with a given netdev. 1343 * Called with a reference on the netdevice and devlink locks only, 1344 * rtnl_lock is not held. 1345 * int (*ndo_tunnel_ctl)(struct net_device *dev, struct ip_tunnel_parm *p, 1346 * int cmd); 1347 * Add, change, delete or get information on an IPv4 tunnel. 1348 * struct net_device *(*ndo_get_peer_dev)(struct net_device *dev); 1349 * If a device is paired with a peer device, return the peer instance. 1350 * The caller must be under RCU read context. 1351 * int (*ndo_fill_forward_path)(struct net_device_path_ctx *ctx, struct net_device_path *path); 1352 * Get the forwarding path to reach the real device from the HW destination address 1353 */ 1354 struct net_device_ops { 1355 int (*ndo_init)(struct net_device *dev); 1356 void (*ndo_uninit)(struct net_device *dev); 1357 int (*ndo_open)(struct net_device *dev); 1358 int (*ndo_stop)(struct net_device *dev); 1359 netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb, 1360 struct net_device *dev); 1361 netdev_features_t (*ndo_features_check)(struct sk_buff *skb, 1362 struct net_device *dev, 1363 netdev_features_t features); 1364 u16 (*ndo_select_queue)(struct net_device *dev, 1365 struct sk_buff *skb, 1366 struct net_device *sb_dev); 1367 void (*ndo_change_rx_flags)(struct net_device *dev, 1368 int flags); 1369 void (*ndo_set_rx_mode)(struct net_device *dev); 1370 int (*ndo_set_mac_address)(struct net_device *dev, 1371 void *addr); 1372 int (*ndo_validate_addr)(struct net_device *dev); 1373 int (*ndo_do_ioctl)(struct net_device *dev, 1374 struct ifreq *ifr, int cmd); 1375 int (*ndo_eth_ioctl)(struct net_device *dev, 1376 struct ifreq *ifr, int cmd); 1377 int (*ndo_siocbond)(struct net_device *dev, 1378 struct ifreq *ifr, int cmd); 1379 int (*ndo_siocwandev)(struct net_device *dev, 1380 struct if_settings *ifs); 1381 int (*ndo_siocdevprivate)(struct net_device *dev, 1382 struct ifreq *ifr, 1383 void __user *data, int cmd); 1384 int (*ndo_set_config)(struct net_device *dev, 1385 struct ifmap *map); 1386 int (*ndo_change_mtu)(struct net_device *dev, 1387 int new_mtu); 1388 int (*ndo_neigh_setup)(struct net_device *dev, 1389 struct neigh_parms *); 1390 void (*ndo_tx_timeout) (struct net_device *dev, 1391 unsigned int txqueue); 1392 1393 void (*ndo_get_stats64)(struct net_device *dev, 1394 struct rtnl_link_stats64 *storage); 1395 bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id); 1396 int (*ndo_get_offload_stats)(int attr_id, 1397 const struct net_device *dev, 1398 void *attr_data); 1399 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev); 1400 1401 int (*ndo_vlan_rx_add_vid)(struct net_device *dev, 1402 __be16 proto, u16 vid); 1403 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, 1404 __be16 proto, u16 vid); 1405 #ifdef CONFIG_NET_POLL_CONTROLLER 1406 void (*ndo_poll_controller)(struct net_device *dev); 1407 int (*ndo_netpoll_setup)(struct net_device *dev, 1408 struct netpoll_info *info); 1409 void (*ndo_netpoll_cleanup)(struct net_device *dev); 1410 #endif 1411 int (*ndo_set_vf_mac)(struct net_device *dev, 1412 int queue, u8 *mac); 1413 int (*ndo_set_vf_vlan)(struct net_device *dev, 1414 int queue, u16 vlan, 1415 u8 qos, __be16 proto); 1416 int (*ndo_set_vf_rate)(struct net_device *dev, 1417 int vf, int min_tx_rate, 1418 int max_tx_rate); 1419 int (*ndo_set_vf_spoofchk)(struct net_device *dev, 1420 int vf, bool setting); 1421 int (*ndo_set_vf_trust)(struct net_device *dev, 1422 int vf, bool setting); 1423 int (*ndo_get_vf_config)(struct net_device *dev, 1424 int vf, 1425 struct ifla_vf_info *ivf); 1426 int (*ndo_set_vf_link_state)(struct net_device *dev, 1427 int vf, int link_state); 1428 int (*ndo_get_vf_stats)(struct net_device *dev, 1429 int vf, 1430 struct ifla_vf_stats 1431 *vf_stats); 1432 int (*ndo_set_vf_port)(struct net_device *dev, 1433 int vf, 1434 struct nlattr *port[]); 1435 int (*ndo_get_vf_port)(struct net_device *dev, 1436 int vf, struct sk_buff *skb); 1437 int (*ndo_get_vf_guid)(struct net_device *dev, 1438 int vf, 1439 struct ifla_vf_guid *node_guid, 1440 struct ifla_vf_guid *port_guid); 1441 int (*ndo_set_vf_guid)(struct net_device *dev, 1442 int vf, u64 guid, 1443 int guid_type); 1444 int (*ndo_set_vf_rss_query_en)( 1445 struct net_device *dev, 1446 int vf, bool setting); 1447 int (*ndo_setup_tc)(struct net_device *dev, 1448 enum tc_setup_type type, 1449 void *type_data); 1450 #if IS_ENABLED(CONFIG_FCOE) 1451 int (*ndo_fcoe_enable)(struct net_device *dev); 1452 int (*ndo_fcoe_disable)(struct net_device *dev); 1453 int (*ndo_fcoe_ddp_setup)(struct net_device *dev, 1454 u16 xid, 1455 struct scatterlist *sgl, 1456 unsigned int sgc); 1457 int (*ndo_fcoe_ddp_done)(struct net_device *dev, 1458 u16 xid); 1459 int (*ndo_fcoe_ddp_target)(struct net_device *dev, 1460 u16 xid, 1461 struct scatterlist *sgl, 1462 unsigned int sgc); 1463 int (*ndo_fcoe_get_hbainfo)(struct net_device *dev, 1464 struct netdev_fcoe_hbainfo *hbainfo); 1465 #endif 1466 1467 #if IS_ENABLED(CONFIG_LIBFCOE) 1468 #define NETDEV_FCOE_WWNN 0 1469 #define NETDEV_FCOE_WWPN 1 1470 int (*ndo_fcoe_get_wwn)(struct net_device *dev, 1471 u64 *wwn, int type); 1472 #endif 1473 1474 #ifdef CONFIG_RFS_ACCEL 1475 int (*ndo_rx_flow_steer)(struct net_device *dev, 1476 const struct sk_buff *skb, 1477 u16 rxq_index, 1478 u32 flow_id); 1479 #endif 1480 int (*ndo_add_slave)(struct net_device *dev, 1481 struct net_device *slave_dev, 1482 struct netlink_ext_ack *extack); 1483 int (*ndo_del_slave)(struct net_device *dev, 1484 struct net_device *slave_dev); 1485 struct net_device* (*ndo_get_xmit_slave)(struct net_device *dev, 1486 struct sk_buff *skb, 1487 bool all_slaves); 1488 struct net_device* (*ndo_sk_get_lower_dev)(struct net_device *dev, 1489 struct sock *sk); 1490 netdev_features_t (*ndo_fix_features)(struct net_device *dev, 1491 netdev_features_t features); 1492 int (*ndo_set_features)(struct net_device *dev, 1493 netdev_features_t features); 1494 int (*ndo_neigh_construct)(struct net_device *dev, 1495 struct neighbour *n); 1496 void (*ndo_neigh_destroy)(struct net_device *dev, 1497 struct neighbour *n); 1498 1499 int (*ndo_fdb_add)(struct ndmsg *ndm, 1500 struct nlattr *tb[], 1501 struct net_device *dev, 1502 const unsigned char *addr, 1503 u16 vid, 1504 u16 flags, 1505 struct netlink_ext_ack *extack); 1506 int (*ndo_fdb_del)(struct ndmsg *ndm, 1507 struct nlattr *tb[], 1508 struct net_device *dev, 1509 const unsigned char *addr, 1510 u16 vid); 1511 int (*ndo_fdb_dump)(struct sk_buff *skb, 1512 struct netlink_callback *cb, 1513 struct net_device *dev, 1514 struct net_device *filter_dev, 1515 int *idx); 1516 int (*ndo_fdb_get)(struct sk_buff *skb, 1517 struct nlattr *tb[], 1518 struct net_device *dev, 1519 const unsigned char *addr, 1520 u16 vid, u32 portid, u32 seq, 1521 struct netlink_ext_ack *extack); 1522 int (*ndo_bridge_setlink)(struct net_device *dev, 1523 struct nlmsghdr *nlh, 1524 u16 flags, 1525 struct netlink_ext_ack *extack); 1526 int (*ndo_bridge_getlink)(struct sk_buff *skb, 1527 u32 pid, u32 seq, 1528 struct net_device *dev, 1529 u32 filter_mask, 1530 int nlflags); 1531 int (*ndo_bridge_dellink)(struct net_device *dev, 1532 struct nlmsghdr *nlh, 1533 u16 flags); 1534 int (*ndo_change_carrier)(struct net_device *dev, 1535 bool new_carrier); 1536 int (*ndo_get_phys_port_id)(struct net_device *dev, 1537 struct netdev_phys_item_id *ppid); 1538 int (*ndo_get_port_parent_id)(struct net_device *dev, 1539 struct netdev_phys_item_id *ppid); 1540 int (*ndo_get_phys_port_name)(struct net_device *dev, 1541 char *name, size_t len); 1542 void* (*ndo_dfwd_add_station)(struct net_device *pdev, 1543 struct net_device *dev); 1544 void (*ndo_dfwd_del_station)(struct net_device *pdev, 1545 void *priv); 1546 1547 int (*ndo_set_tx_maxrate)(struct net_device *dev, 1548 int queue_index, 1549 u32 maxrate); 1550 int (*ndo_get_iflink)(const struct net_device *dev); 1551 int (*ndo_fill_metadata_dst)(struct net_device *dev, 1552 struct sk_buff *skb); 1553 void (*ndo_set_rx_headroom)(struct net_device *dev, 1554 int needed_headroom); 1555 int (*ndo_bpf)(struct net_device *dev, 1556 struct netdev_bpf *bpf); 1557 int (*ndo_xdp_xmit)(struct net_device *dev, int n, 1558 struct xdp_frame **xdp, 1559 u32 flags); 1560 struct net_device * (*ndo_xdp_get_xmit_slave)(struct net_device *dev, 1561 struct xdp_buff *xdp); 1562 int (*ndo_xsk_wakeup)(struct net_device *dev, 1563 u32 queue_id, u32 flags); 1564 struct devlink_port * (*ndo_get_devlink_port)(struct net_device *dev); 1565 int (*ndo_tunnel_ctl)(struct net_device *dev, 1566 struct ip_tunnel_parm *p, int cmd); 1567 struct net_device * (*ndo_get_peer_dev)(struct net_device *dev); 1568 int (*ndo_fill_forward_path)(struct net_device_path_ctx *ctx, 1569 struct net_device_path *path); 1570 }; 1571 1572 /** 1573 * enum netdev_priv_flags - &struct net_device priv_flags 1574 * 1575 * These are the &struct net_device, they are only set internally 1576 * by drivers and used in the kernel. These flags are invisible to 1577 * userspace; this means that the order of these flags can change 1578 * during any kernel release. 1579 * 1580 * You should have a pretty good reason to be extending these flags. 1581 * 1582 * @IFF_802_1Q_VLAN: 802.1Q VLAN device 1583 * @IFF_EBRIDGE: Ethernet bridging device 1584 * @IFF_BONDING: bonding master or slave 1585 * @IFF_ISATAP: ISATAP interface (RFC4214) 1586 * @IFF_WAN_HDLC: WAN HDLC device 1587 * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to 1588 * release skb->dst 1589 * @IFF_DONT_BRIDGE: disallow bridging this ether dev 1590 * @IFF_DISABLE_NETPOLL: disable netpoll at run-time 1591 * @IFF_MACVLAN_PORT: device used as macvlan port 1592 * @IFF_BRIDGE_PORT: device used as bridge port 1593 * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port 1594 * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit 1595 * @IFF_UNICAST_FLT: Supports unicast filtering 1596 * @IFF_TEAM_PORT: device used as team port 1597 * @IFF_SUPP_NOFCS: device supports sending custom FCS 1598 * @IFF_LIVE_ADDR_CHANGE: device supports hardware address 1599 * change when it's running 1600 * @IFF_MACVLAN: Macvlan device 1601 * @IFF_XMIT_DST_RELEASE_PERM: IFF_XMIT_DST_RELEASE not taking into account 1602 * underlying stacked devices 1603 * @IFF_L3MDEV_MASTER: device is an L3 master device 1604 * @IFF_NO_QUEUE: device can run without qdisc attached 1605 * @IFF_OPENVSWITCH: device is a Open vSwitch master 1606 * @IFF_L3MDEV_SLAVE: device is enslaved to an L3 master device 1607 * @IFF_TEAM: device is a team device 1608 * @IFF_RXFH_CONFIGURED: device has had Rx Flow indirection table configured 1609 * @IFF_PHONY_HEADROOM: the headroom value is controlled by an external 1610 * entity (i.e. the master device for bridged veth) 1611 * @IFF_MACSEC: device is a MACsec device 1612 * @IFF_NO_RX_HANDLER: device doesn't support the rx_handler hook 1613 * @IFF_FAILOVER: device is a failover master device 1614 * @IFF_FAILOVER_SLAVE: device is lower dev of a failover master device 1615 * @IFF_L3MDEV_RX_HANDLER: only invoke the rx handler of L3 master device 1616 * @IFF_LIVE_RENAME_OK: rename is allowed while device is up and running 1617 * @IFF_TX_SKB_NO_LINEAR: device/driver is capable of xmitting frames with 1618 * skb_headlen(skb) == 0 (data starts from frag0) 1619 * @IFF_CHANGE_PROTO_DOWN: device supports setting carrier via IFLA_PROTO_DOWN 1620 */ 1621 enum netdev_priv_flags { 1622 IFF_802_1Q_VLAN = 1<<0, 1623 IFF_EBRIDGE = 1<<1, 1624 IFF_BONDING = 1<<2, 1625 IFF_ISATAP = 1<<3, 1626 IFF_WAN_HDLC = 1<<4, 1627 IFF_XMIT_DST_RELEASE = 1<<5, 1628 IFF_DONT_BRIDGE = 1<<6, 1629 IFF_DISABLE_NETPOLL = 1<<7, 1630 IFF_MACVLAN_PORT = 1<<8, 1631 IFF_BRIDGE_PORT = 1<<9, 1632 IFF_OVS_DATAPATH = 1<<10, 1633 IFF_TX_SKB_SHARING = 1<<11, 1634 IFF_UNICAST_FLT = 1<<12, 1635 IFF_TEAM_PORT = 1<<13, 1636 IFF_SUPP_NOFCS = 1<<14, 1637 IFF_LIVE_ADDR_CHANGE = 1<<15, 1638 IFF_MACVLAN = 1<<16, 1639 IFF_XMIT_DST_RELEASE_PERM = 1<<17, 1640 IFF_L3MDEV_MASTER = 1<<18, 1641 IFF_NO_QUEUE = 1<<19, 1642 IFF_OPENVSWITCH = 1<<20, 1643 IFF_L3MDEV_SLAVE = 1<<21, 1644 IFF_TEAM = 1<<22, 1645 IFF_RXFH_CONFIGURED = 1<<23, 1646 IFF_PHONY_HEADROOM = 1<<24, 1647 IFF_MACSEC = 1<<25, 1648 IFF_NO_RX_HANDLER = 1<<26, 1649 IFF_FAILOVER = 1<<27, 1650 IFF_FAILOVER_SLAVE = 1<<28, 1651 IFF_L3MDEV_RX_HANDLER = 1<<29, 1652 IFF_LIVE_RENAME_OK = 1<<30, 1653 IFF_TX_SKB_NO_LINEAR = 1<<31, 1654 IFF_CHANGE_PROTO_DOWN = BIT_ULL(32), 1655 }; 1656 1657 #define IFF_802_1Q_VLAN IFF_802_1Q_VLAN 1658 #define IFF_EBRIDGE IFF_EBRIDGE 1659 #define IFF_BONDING IFF_BONDING 1660 #define IFF_ISATAP IFF_ISATAP 1661 #define IFF_WAN_HDLC IFF_WAN_HDLC 1662 #define IFF_XMIT_DST_RELEASE IFF_XMIT_DST_RELEASE 1663 #define IFF_DONT_BRIDGE IFF_DONT_BRIDGE 1664 #define IFF_DISABLE_NETPOLL IFF_DISABLE_NETPOLL 1665 #define IFF_MACVLAN_PORT IFF_MACVLAN_PORT 1666 #define IFF_BRIDGE_PORT IFF_BRIDGE_PORT 1667 #define IFF_OVS_DATAPATH IFF_OVS_DATAPATH 1668 #define IFF_TX_SKB_SHARING IFF_TX_SKB_SHARING 1669 #define IFF_UNICAST_FLT IFF_UNICAST_FLT 1670 #define IFF_TEAM_PORT IFF_TEAM_PORT 1671 #define IFF_SUPP_NOFCS IFF_SUPP_NOFCS 1672 #define IFF_LIVE_ADDR_CHANGE IFF_LIVE_ADDR_CHANGE 1673 #define IFF_MACVLAN IFF_MACVLAN 1674 #define IFF_XMIT_DST_RELEASE_PERM IFF_XMIT_DST_RELEASE_PERM 1675 #define IFF_L3MDEV_MASTER IFF_L3MDEV_MASTER 1676 #define IFF_NO_QUEUE IFF_NO_QUEUE 1677 #define IFF_OPENVSWITCH IFF_OPENVSWITCH 1678 #define IFF_L3MDEV_SLAVE IFF_L3MDEV_SLAVE 1679 #define IFF_TEAM IFF_TEAM 1680 #define IFF_RXFH_CONFIGURED IFF_RXFH_CONFIGURED 1681 #define IFF_PHONY_HEADROOM IFF_PHONY_HEADROOM 1682 #define IFF_MACSEC IFF_MACSEC 1683 #define IFF_NO_RX_HANDLER IFF_NO_RX_HANDLER 1684 #define IFF_FAILOVER IFF_FAILOVER 1685 #define IFF_FAILOVER_SLAVE IFF_FAILOVER_SLAVE 1686 #define IFF_L3MDEV_RX_HANDLER IFF_L3MDEV_RX_HANDLER 1687 #define IFF_LIVE_RENAME_OK IFF_LIVE_RENAME_OK 1688 #define IFF_TX_SKB_NO_LINEAR IFF_TX_SKB_NO_LINEAR 1689 1690 /* Specifies the type of the struct net_device::ml_priv pointer */ 1691 enum netdev_ml_priv_type { 1692 ML_PRIV_NONE, 1693 ML_PRIV_CAN, 1694 }; 1695 1696 /** 1697 * struct net_device - The DEVICE structure. 1698 * 1699 * Actually, this whole structure is a big mistake. It mixes I/O 1700 * data with strictly "high-level" data, and it has to know about 1701 * almost every data structure used in the INET module. 1702 * 1703 * @name: This is the first field of the "visible" part of this structure 1704 * (i.e. as seen by users in the "Space.c" file). It is the name 1705 * of the interface. 1706 * 1707 * @name_node: Name hashlist node 1708 * @ifalias: SNMP alias 1709 * @mem_end: Shared memory end 1710 * @mem_start: Shared memory start 1711 * @base_addr: Device I/O address 1712 * @irq: Device IRQ number 1713 * 1714 * @state: Generic network queuing layer state, see netdev_state_t 1715 * @dev_list: The global list of network devices 1716 * @napi_list: List entry used for polling NAPI devices 1717 * @unreg_list: List entry when we are unregistering the 1718 * device; see the function unregister_netdev 1719 * @close_list: List entry used when we are closing the device 1720 * @ptype_all: Device-specific packet handlers for all protocols 1721 * @ptype_specific: Device-specific, protocol-specific packet handlers 1722 * 1723 * @adj_list: Directly linked devices, like slaves for bonding 1724 * @features: Currently active device features 1725 * @hw_features: User-changeable features 1726 * 1727 * @wanted_features: User-requested features 1728 * @vlan_features: Mask of features inheritable by VLAN devices 1729 * 1730 * @hw_enc_features: Mask of features inherited by encapsulating devices 1731 * This field indicates what encapsulation 1732 * offloads the hardware is capable of doing, 1733 * and drivers will need to set them appropriately. 1734 * 1735 * @mpls_features: Mask of features inheritable by MPLS 1736 * @gso_partial_features: value(s) from NETIF_F_GSO\* 1737 * 1738 * @ifindex: interface index 1739 * @group: The group the device belongs to 1740 * 1741 * @stats: Statistics struct, which was left as a legacy, use 1742 * rtnl_link_stats64 instead 1743 * 1744 * @rx_dropped: Dropped packets by core network, 1745 * do not use this in drivers 1746 * @tx_dropped: Dropped packets by core network, 1747 * do not use this in drivers 1748 * @rx_nohandler: nohandler dropped packets by core network on 1749 * inactive devices, do not use this in drivers 1750 * @carrier_up_count: Number of times the carrier has been up 1751 * @carrier_down_count: Number of times the carrier has been down 1752 * 1753 * @wireless_handlers: List of functions to handle Wireless Extensions, 1754 * instead of ioctl, 1755 * see <net/iw_handler.h> for details. 1756 * @wireless_data: Instance data managed by the core of wireless extensions 1757 * 1758 * @netdev_ops: Includes several pointers to callbacks, 1759 * if one wants to override the ndo_*() functions 1760 * @ethtool_ops: Management operations 1761 * @l3mdev_ops: Layer 3 master device operations 1762 * @ndisc_ops: Includes callbacks for different IPv6 neighbour 1763 * discovery handling. Necessary for e.g. 6LoWPAN. 1764 * @xfrmdev_ops: Transformation offload operations 1765 * @tlsdev_ops: Transport Layer Security offload operations 1766 * @header_ops: Includes callbacks for creating,parsing,caching,etc 1767 * of Layer 2 headers. 1768 * 1769 * @flags: Interface flags (a la BSD) 1770 * @priv_flags: Like 'flags' but invisible to userspace, 1771 * see if.h for the definitions 1772 * @gflags: Global flags ( kept as legacy ) 1773 * @padded: How much padding added by alloc_netdev() 1774 * @operstate: RFC2863 operstate 1775 * @link_mode: Mapping policy to operstate 1776 * @if_port: Selectable AUI, TP, ... 1777 * @dma: DMA channel 1778 * @mtu: Interface MTU value 1779 * @min_mtu: Interface Minimum MTU value 1780 * @max_mtu: Interface Maximum MTU value 1781 * @type: Interface hardware type 1782 * @hard_header_len: Maximum hardware header length. 1783 * @min_header_len: Minimum hardware header length 1784 * 1785 * @needed_headroom: Extra headroom the hardware may need, but not in all 1786 * cases can this be guaranteed 1787 * @needed_tailroom: Extra tailroom the hardware may need, but not in all 1788 * cases can this be guaranteed. Some cases also use 1789 * LL_MAX_HEADER instead to allocate the skb 1790 * 1791 * interface address info: 1792 * 1793 * @perm_addr: Permanent hw address 1794 * @addr_assign_type: Hw address assignment type 1795 * @addr_len: Hardware address length 1796 * @upper_level: Maximum depth level of upper devices. 1797 * @lower_level: Maximum depth level of lower devices. 1798 * @neigh_priv_len: Used in neigh_alloc() 1799 * @dev_id: Used to differentiate devices that share 1800 * the same link layer address 1801 * @dev_port: Used to differentiate devices that share 1802 * the same function 1803 * @addr_list_lock: XXX: need comments on this one 1804 * @name_assign_type: network interface name assignment type 1805 * @uc_promisc: Counter that indicates promiscuous mode 1806 * has been enabled due to the need to listen to 1807 * additional unicast addresses in a device that 1808 * does not implement ndo_set_rx_mode() 1809 * @uc: unicast mac addresses 1810 * @mc: multicast mac addresses 1811 * @dev_addrs: list of device hw addresses 1812 * @queues_kset: Group of all Kobjects in the Tx and RX queues 1813 * @promiscuity: Number of times the NIC is told to work in 1814 * promiscuous mode; if it becomes 0 the NIC will 1815 * exit promiscuous mode 1816 * @allmulti: Counter, enables or disables allmulticast mode 1817 * 1818 * @vlan_info: VLAN info 1819 * @dsa_ptr: dsa specific data 1820 * @tipc_ptr: TIPC specific data 1821 * @atalk_ptr: AppleTalk link 1822 * @ip_ptr: IPv4 specific data 1823 * @dn_ptr: DECnet specific data 1824 * @ip6_ptr: IPv6 specific data 1825 * @ax25_ptr: AX.25 specific data 1826 * @ieee80211_ptr: IEEE 802.11 specific data, assign before registering 1827 * @ieee802154_ptr: IEEE 802.15.4 low-rate Wireless Personal Area Network 1828 * device struct 1829 * @mpls_ptr: mpls_dev struct pointer 1830 * @mctp_ptr: MCTP specific data 1831 * 1832 * @dev_addr: Hw address (before bcast, 1833 * because most packets are unicast) 1834 * 1835 * @_rx: Array of RX queues 1836 * @num_rx_queues: Number of RX queues 1837 * allocated at register_netdev() time 1838 * @real_num_rx_queues: Number of RX queues currently active in device 1839 * @xdp_prog: XDP sockets filter program pointer 1840 * @gro_flush_timeout: timeout for GRO layer in NAPI 1841 * @napi_defer_hard_irqs: If not zero, provides a counter that would 1842 * allow to avoid NIC hard IRQ, on busy queues. 1843 * 1844 * @rx_handler: handler for received packets 1845 * @rx_handler_data: XXX: need comments on this one 1846 * @miniq_ingress: ingress/clsact qdisc specific data for 1847 * ingress processing 1848 * @ingress_queue: XXX: need comments on this one 1849 * @nf_hooks_ingress: netfilter hooks executed for ingress packets 1850 * @broadcast: hw bcast address 1851 * 1852 * @rx_cpu_rmap: CPU reverse-mapping for RX completion interrupts, 1853 * indexed by RX queue number. Assigned by driver. 1854 * This must only be set if the ndo_rx_flow_steer 1855 * operation is defined 1856 * @index_hlist: Device index hash chain 1857 * 1858 * @_tx: Array of TX queues 1859 * @num_tx_queues: Number of TX queues allocated at alloc_netdev_mq() time 1860 * @real_num_tx_queues: Number of TX queues currently active in device 1861 * @qdisc: Root qdisc from userspace point of view 1862 * @tx_queue_len: Max frames per queue allowed 1863 * @tx_global_lock: XXX: need comments on this one 1864 * @xdp_bulkq: XDP device bulk queue 1865 * @xps_maps: all CPUs/RXQs maps for XPS device 1866 * 1867 * @xps_maps: XXX: need comments on this one 1868 * @miniq_egress: clsact qdisc specific data for 1869 * egress processing 1870 * @nf_hooks_egress: netfilter hooks executed for egress packets 1871 * @qdisc_hash: qdisc hash table 1872 * @watchdog_timeo: Represents the timeout that is used by 1873 * the watchdog (see dev_watchdog()) 1874 * @watchdog_timer: List of timers 1875 * 1876 * @proto_down_reason: reason a netdev interface is held down 1877 * @pcpu_refcnt: Number of references to this device 1878 * @dev_refcnt: Number of references to this device 1879 * @refcnt_tracker: Tracker directory for tracked references to this device 1880 * @todo_list: Delayed register/unregister 1881 * @link_watch_list: XXX: need comments on this one 1882 * 1883 * @reg_state: Register/unregister state machine 1884 * @dismantle: Device is going to be freed 1885 * @rtnl_link_state: This enum represents the phases of creating 1886 * a new link 1887 * 1888 * @needs_free_netdev: Should unregister perform free_netdev? 1889 * @priv_destructor: Called from unregister 1890 * @npinfo: XXX: need comments on this one 1891 * @nd_net: Network namespace this network device is inside 1892 * 1893 * @ml_priv: Mid-layer private 1894 * @ml_priv_type: Mid-layer private type 1895 * @lstats: Loopback statistics 1896 * @tstats: Tunnel statistics 1897 * @dstats: Dummy statistics 1898 * @vstats: Virtual ethernet statistics 1899 * 1900 * @garp_port: GARP 1901 * @mrp_port: MRP 1902 * 1903 * @dev: Class/net/name entry 1904 * @sysfs_groups: Space for optional device, statistics and wireless 1905 * sysfs groups 1906 * 1907 * @sysfs_rx_queue_group: Space for optional per-rx queue attributes 1908 * @rtnl_link_ops: Rtnl_link_ops 1909 * 1910 * @gso_max_size: Maximum size of generic segmentation offload 1911 * @gso_max_segs: Maximum number of segments that can be passed to the 1912 * NIC for GSO 1913 * 1914 * @dcbnl_ops: Data Center Bridging netlink ops 1915 * @num_tc: Number of traffic classes in the net device 1916 * @tc_to_txq: XXX: need comments on this one 1917 * @prio_tc_map: XXX: need comments on this one 1918 * 1919 * @fcoe_ddp_xid: Max exchange id for FCoE LRO by ddp 1920 * 1921 * @priomap: XXX: need comments on this one 1922 * @phydev: Physical device may attach itself 1923 * for hardware timestamping 1924 * @sfp_bus: attached &struct sfp_bus structure. 1925 * 1926 * @qdisc_tx_busylock: lockdep class annotating Qdisc->busylock spinlock 1927 * 1928 * @proto_down: protocol port state information can be sent to the 1929 * switch driver and used to set the phys state of the 1930 * switch port. 1931 * 1932 * @wol_enabled: Wake-on-LAN is enabled 1933 * 1934 * @threaded: napi threaded mode is enabled 1935 * 1936 * @net_notifier_list: List of per-net netdev notifier block 1937 * that follow this device when it is moved 1938 * to another network namespace. 1939 * 1940 * @macsec_ops: MACsec offloading ops 1941 * 1942 * @udp_tunnel_nic_info: static structure describing the UDP tunnel 1943 * offload capabilities of the device 1944 * @udp_tunnel_nic: UDP tunnel offload state 1945 * @xdp_state: stores info on attached XDP BPF programs 1946 * 1947 * @nested_level: Used as as a parameter of spin_lock_nested() of 1948 * dev->addr_list_lock. 1949 * @unlink_list: As netif_addr_lock() can be called recursively, 1950 * keep a list of interfaces to be deleted. 1951 * 1952 * @dev_addr_shadow: Copy of @dev_addr to catch direct writes. 1953 * @linkwatch_dev_tracker: refcount tracker used by linkwatch. 1954 * @watchdog_dev_tracker: refcount tracker used by watchdog. 1955 * 1956 * FIXME: cleanup struct net_device such that network protocol info 1957 * moves out. 1958 */ 1959 1960 struct net_device { 1961 char name[IFNAMSIZ]; 1962 struct netdev_name_node *name_node; 1963 struct dev_ifalias __rcu *ifalias; 1964 /* 1965 * I/O specific fields 1966 * FIXME: Merge these and struct ifmap into one 1967 */ 1968 unsigned long mem_end; 1969 unsigned long mem_start; 1970 unsigned long base_addr; 1971 1972 /* 1973 * Some hardware also needs these fields (state,dev_list, 1974 * napi_list,unreg_list,close_list) but they are not 1975 * part of the usual set specified in Space.c. 1976 */ 1977 1978 unsigned long state; 1979 1980 struct list_head dev_list; 1981 struct list_head napi_list; 1982 struct list_head unreg_list; 1983 struct list_head close_list; 1984 struct list_head ptype_all; 1985 struct list_head ptype_specific; 1986 1987 struct { 1988 struct list_head upper; 1989 struct list_head lower; 1990 } adj_list; 1991 1992 /* Read-mostly cache-line for fast-path access */ 1993 unsigned int flags; 1994 unsigned long long priv_flags; 1995 const struct net_device_ops *netdev_ops; 1996 int ifindex; 1997 unsigned short gflags; 1998 unsigned short hard_header_len; 1999 2000 /* Note : dev->mtu is often read without holding a lock. 2001 * Writers usually hold RTNL. 2002 * It is recommended to use READ_ONCE() to annotate the reads, 2003 * and to use WRITE_ONCE() to annotate the writes. 2004 */ 2005 unsigned int mtu; 2006 unsigned short needed_headroom; 2007 unsigned short needed_tailroom; 2008 2009 netdev_features_t features; 2010 netdev_features_t hw_features; 2011 netdev_features_t wanted_features; 2012 netdev_features_t vlan_features; 2013 netdev_features_t hw_enc_features; 2014 netdev_features_t mpls_features; 2015 netdev_features_t gso_partial_features; 2016 2017 unsigned int min_mtu; 2018 unsigned int max_mtu; 2019 unsigned short type; 2020 unsigned char min_header_len; 2021 unsigned char name_assign_type; 2022 2023 int group; 2024 2025 struct net_device_stats stats; /* not used by modern drivers */ 2026 2027 atomic_long_t rx_dropped; 2028 atomic_long_t tx_dropped; 2029 atomic_long_t rx_nohandler; 2030 2031 /* Stats to monitor link on/off, flapping */ 2032 atomic_t carrier_up_count; 2033 atomic_t carrier_down_count; 2034 2035 #ifdef CONFIG_WIRELESS_EXT 2036 const struct iw_handler_def *wireless_handlers; 2037 struct iw_public_data *wireless_data; 2038 #endif 2039 const struct ethtool_ops *ethtool_ops; 2040 #ifdef CONFIG_NET_L3_MASTER_DEV 2041 const struct l3mdev_ops *l3mdev_ops; 2042 #endif 2043 #if IS_ENABLED(CONFIG_IPV6) 2044 const struct ndisc_ops *ndisc_ops; 2045 #endif 2046 2047 #ifdef CONFIG_XFRM_OFFLOAD 2048 const struct xfrmdev_ops *xfrmdev_ops; 2049 #endif 2050 2051 #if IS_ENABLED(CONFIG_TLS_DEVICE) 2052 const struct tlsdev_ops *tlsdev_ops; 2053 #endif 2054 2055 const struct header_ops *header_ops; 2056 2057 unsigned char operstate; 2058 unsigned char link_mode; 2059 2060 unsigned char if_port; 2061 unsigned char dma; 2062 2063 /* Interface address info. */ 2064 unsigned char perm_addr[MAX_ADDR_LEN]; 2065 unsigned char addr_assign_type; 2066 unsigned char addr_len; 2067 unsigned char upper_level; 2068 unsigned char lower_level; 2069 2070 unsigned short neigh_priv_len; 2071 unsigned short dev_id; 2072 unsigned short dev_port; 2073 unsigned short padded; 2074 2075 spinlock_t addr_list_lock; 2076 int irq; 2077 2078 struct netdev_hw_addr_list uc; 2079 struct netdev_hw_addr_list mc; 2080 struct netdev_hw_addr_list dev_addrs; 2081 2082 #ifdef CONFIG_SYSFS 2083 struct kset *queues_kset; 2084 #endif 2085 #ifdef CONFIG_LOCKDEP 2086 struct list_head unlink_list; 2087 #endif 2088 unsigned int promiscuity; 2089 unsigned int allmulti; 2090 bool uc_promisc; 2091 #ifdef CONFIG_LOCKDEP 2092 unsigned char nested_level; 2093 #endif 2094 2095 2096 /* Protocol-specific pointers */ 2097 2098 #if IS_ENABLED(CONFIG_VLAN_8021Q) 2099 struct vlan_info __rcu *vlan_info; 2100 #endif 2101 #if IS_ENABLED(CONFIG_NET_DSA) 2102 struct dsa_port *dsa_ptr; 2103 #endif 2104 #if IS_ENABLED(CONFIG_TIPC) 2105 struct tipc_bearer __rcu *tipc_ptr; 2106 #endif 2107 #if IS_ENABLED(CONFIG_IRDA) || IS_ENABLED(CONFIG_ATALK) 2108 void *atalk_ptr; 2109 #endif 2110 struct in_device __rcu *ip_ptr; 2111 #if IS_ENABLED(CONFIG_DECNET) 2112 struct dn_dev __rcu *dn_ptr; 2113 #endif 2114 struct inet6_dev __rcu *ip6_ptr; 2115 #if IS_ENABLED(CONFIG_AX25) 2116 void *ax25_ptr; 2117 #endif 2118 struct wireless_dev *ieee80211_ptr; 2119 struct wpan_dev *ieee802154_ptr; 2120 #if IS_ENABLED(CONFIG_MPLS_ROUTING) 2121 struct mpls_dev __rcu *mpls_ptr; 2122 #endif 2123 #if IS_ENABLED(CONFIG_MCTP) 2124 struct mctp_dev __rcu *mctp_ptr; 2125 #endif 2126 2127 /* 2128 * Cache lines mostly used on receive path (including eth_type_trans()) 2129 */ 2130 /* Interface address info used in eth_type_trans() */ 2131 const unsigned char *dev_addr; 2132 2133 struct netdev_rx_queue *_rx; 2134 unsigned int num_rx_queues; 2135 unsigned int real_num_rx_queues; 2136 2137 struct bpf_prog __rcu *xdp_prog; 2138 unsigned long gro_flush_timeout; 2139 int napi_defer_hard_irqs; 2140 rx_handler_func_t __rcu *rx_handler; 2141 void __rcu *rx_handler_data; 2142 2143 #ifdef CONFIG_NET_CLS_ACT 2144 struct mini_Qdisc __rcu *miniq_ingress; 2145 #endif 2146 struct netdev_queue __rcu *ingress_queue; 2147 #ifdef CONFIG_NETFILTER_INGRESS 2148 struct nf_hook_entries __rcu *nf_hooks_ingress; 2149 #endif 2150 2151 unsigned char broadcast[MAX_ADDR_LEN]; 2152 #ifdef CONFIG_RFS_ACCEL 2153 struct cpu_rmap *rx_cpu_rmap; 2154 #endif 2155 struct hlist_node index_hlist; 2156 2157 /* 2158 * Cache lines mostly used on transmit path 2159 */ 2160 struct netdev_queue *_tx ____cacheline_aligned_in_smp; 2161 unsigned int num_tx_queues; 2162 unsigned int real_num_tx_queues; 2163 struct Qdisc *qdisc; 2164 unsigned int tx_queue_len; 2165 spinlock_t tx_global_lock; 2166 2167 struct xdp_dev_bulk_queue __percpu *xdp_bulkq; 2168 2169 #ifdef CONFIG_XPS 2170 struct xps_dev_maps __rcu *xps_maps[XPS_MAPS_MAX]; 2171 #endif 2172 #ifdef CONFIG_NET_CLS_ACT 2173 struct mini_Qdisc __rcu *miniq_egress; 2174 #endif 2175 #ifdef CONFIG_NETFILTER_EGRESS 2176 struct nf_hook_entries __rcu *nf_hooks_egress; 2177 #endif 2178 2179 #ifdef CONFIG_NET_SCHED 2180 DECLARE_HASHTABLE (qdisc_hash, 4); 2181 #endif 2182 /* These may be needed for future network-power-down code. */ 2183 struct timer_list watchdog_timer; 2184 int watchdog_timeo; 2185 2186 u32 proto_down_reason; 2187 2188 struct list_head todo_list; 2189 2190 #ifdef CONFIG_PCPU_DEV_REFCNT 2191 int __percpu *pcpu_refcnt; 2192 #else 2193 refcount_t dev_refcnt; 2194 #endif 2195 struct ref_tracker_dir refcnt_tracker; 2196 2197 struct list_head link_watch_list; 2198 2199 enum { NETREG_UNINITIALIZED=0, 2200 NETREG_REGISTERED, /* completed register_netdevice */ 2201 NETREG_UNREGISTERING, /* called unregister_netdevice */ 2202 NETREG_UNREGISTERED, /* completed unregister todo */ 2203 NETREG_RELEASED, /* called free_netdev */ 2204 NETREG_DUMMY, /* dummy device for NAPI poll */ 2205 } reg_state:8; 2206 2207 bool dismantle; 2208 2209 enum { 2210 RTNL_LINK_INITIALIZED, 2211 RTNL_LINK_INITIALIZING, 2212 } rtnl_link_state:16; 2213 2214 bool needs_free_netdev; 2215 void (*priv_destructor)(struct net_device *dev); 2216 2217 #ifdef CONFIG_NETPOLL 2218 struct netpoll_info __rcu *npinfo; 2219 #endif 2220 2221 possible_net_t nd_net; 2222 2223 /* mid-layer private */ 2224 void *ml_priv; 2225 enum netdev_ml_priv_type ml_priv_type; 2226 2227 union { 2228 struct pcpu_lstats __percpu *lstats; 2229 struct pcpu_sw_netstats __percpu *tstats; 2230 struct pcpu_dstats __percpu *dstats; 2231 }; 2232 2233 #if IS_ENABLED(CONFIG_GARP) 2234 struct garp_port __rcu *garp_port; 2235 #endif 2236 #if IS_ENABLED(CONFIG_MRP) 2237 struct mrp_port __rcu *mrp_port; 2238 #endif 2239 2240 struct device dev; 2241 const struct attribute_group *sysfs_groups[4]; 2242 const struct attribute_group *sysfs_rx_queue_group; 2243 2244 const struct rtnl_link_ops *rtnl_link_ops; 2245 2246 /* for setting kernel sock attribute on TCP connection setup */ 2247 #define GSO_MAX_SIZE 65536 2248 unsigned int gso_max_size; 2249 #define GSO_MAX_SEGS 65535 2250 u16 gso_max_segs; 2251 2252 #ifdef CONFIG_DCB 2253 const struct dcbnl_rtnl_ops *dcbnl_ops; 2254 #endif 2255 s16 num_tc; 2256 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE]; 2257 u8 prio_tc_map[TC_BITMASK + 1]; 2258 2259 #if IS_ENABLED(CONFIG_FCOE) 2260 unsigned int fcoe_ddp_xid; 2261 #endif 2262 #if IS_ENABLED(CONFIG_CGROUP_NET_PRIO) 2263 struct netprio_map __rcu *priomap; 2264 #endif 2265 struct phy_device *phydev; 2266 struct sfp_bus *sfp_bus; 2267 struct lock_class_key *qdisc_tx_busylock; 2268 bool proto_down; 2269 unsigned wol_enabled:1; 2270 unsigned threaded:1; 2271 2272 struct list_head net_notifier_list; 2273 2274 #if IS_ENABLED(CONFIG_MACSEC) 2275 /* MACsec management functions */ 2276 const struct macsec_ops *macsec_ops; 2277 #endif 2278 const struct udp_tunnel_nic_info *udp_tunnel_nic_info; 2279 struct udp_tunnel_nic *udp_tunnel_nic; 2280 2281 /* protected by rtnl_lock */ 2282 struct bpf_xdp_entity xdp_state[__MAX_XDP_MODE]; 2283 2284 u8 dev_addr_shadow[MAX_ADDR_LEN]; 2285 netdevice_tracker linkwatch_dev_tracker; 2286 netdevice_tracker watchdog_dev_tracker; 2287 }; 2288 #define to_net_dev(d) container_of(d, struct net_device, dev) 2289 2290 static inline bool netif_elide_gro(const struct net_device *dev) 2291 { 2292 if (!(dev->features & NETIF_F_GRO) || dev->xdp_prog) 2293 return true; 2294 return false; 2295 } 2296 2297 #define NETDEV_ALIGN 32 2298 2299 static inline 2300 int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio) 2301 { 2302 return dev->prio_tc_map[prio & TC_BITMASK]; 2303 } 2304 2305 static inline 2306 int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc) 2307 { 2308 if (tc >= dev->num_tc) 2309 return -EINVAL; 2310 2311 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK; 2312 return 0; 2313 } 2314 2315 int netdev_txq_to_tc(struct net_device *dev, unsigned int txq); 2316 void netdev_reset_tc(struct net_device *dev); 2317 int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset); 2318 int netdev_set_num_tc(struct net_device *dev, u8 num_tc); 2319 2320 static inline 2321 int netdev_get_num_tc(struct net_device *dev) 2322 { 2323 return dev->num_tc; 2324 } 2325 2326 static inline void net_prefetch(void *p) 2327 { 2328 prefetch(p); 2329 #if L1_CACHE_BYTES < 128 2330 prefetch((u8 *)p + L1_CACHE_BYTES); 2331 #endif 2332 } 2333 2334 static inline void net_prefetchw(void *p) 2335 { 2336 prefetchw(p); 2337 #if L1_CACHE_BYTES < 128 2338 prefetchw((u8 *)p + L1_CACHE_BYTES); 2339 #endif 2340 } 2341 2342 void netdev_unbind_sb_channel(struct net_device *dev, 2343 struct net_device *sb_dev); 2344 int netdev_bind_sb_channel_queue(struct net_device *dev, 2345 struct net_device *sb_dev, 2346 u8 tc, u16 count, u16 offset); 2347 int netdev_set_sb_channel(struct net_device *dev, u16 channel); 2348 static inline int netdev_get_sb_channel(struct net_device *dev) 2349 { 2350 return max_t(int, -dev->num_tc, 0); 2351 } 2352 2353 static inline 2354 struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev, 2355 unsigned int index) 2356 { 2357 return &dev->_tx[index]; 2358 } 2359 2360 static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev, 2361 const struct sk_buff *skb) 2362 { 2363 return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb)); 2364 } 2365 2366 static inline void netdev_for_each_tx_queue(struct net_device *dev, 2367 void (*f)(struct net_device *, 2368 struct netdev_queue *, 2369 void *), 2370 void *arg) 2371 { 2372 unsigned int i; 2373 2374 for (i = 0; i < dev->num_tx_queues; i++) 2375 f(dev, &dev->_tx[i], arg); 2376 } 2377 2378 #define netdev_lockdep_set_classes(dev) \ 2379 { \ 2380 static struct lock_class_key qdisc_tx_busylock_key; \ 2381 static struct lock_class_key qdisc_xmit_lock_key; \ 2382 static struct lock_class_key dev_addr_list_lock_key; \ 2383 unsigned int i; \ 2384 \ 2385 (dev)->qdisc_tx_busylock = &qdisc_tx_busylock_key; \ 2386 lockdep_set_class(&(dev)->addr_list_lock, \ 2387 &dev_addr_list_lock_key); \ 2388 for (i = 0; i < (dev)->num_tx_queues; i++) \ 2389 lockdep_set_class(&(dev)->_tx[i]._xmit_lock, \ 2390 &qdisc_xmit_lock_key); \ 2391 } 2392 2393 u16 netdev_pick_tx(struct net_device *dev, struct sk_buff *skb, 2394 struct net_device *sb_dev); 2395 struct netdev_queue *netdev_core_pick_tx(struct net_device *dev, 2396 struct sk_buff *skb, 2397 struct net_device *sb_dev); 2398 2399 /* returns the headroom that the master device needs to take in account 2400 * when forwarding to this dev 2401 */ 2402 static inline unsigned netdev_get_fwd_headroom(struct net_device *dev) 2403 { 2404 return dev->priv_flags & IFF_PHONY_HEADROOM ? 0 : dev->needed_headroom; 2405 } 2406 2407 static inline void netdev_set_rx_headroom(struct net_device *dev, int new_hr) 2408 { 2409 if (dev->netdev_ops->ndo_set_rx_headroom) 2410 dev->netdev_ops->ndo_set_rx_headroom(dev, new_hr); 2411 } 2412 2413 /* set the device rx headroom to the dev's default */ 2414 static inline void netdev_reset_rx_headroom(struct net_device *dev) 2415 { 2416 netdev_set_rx_headroom(dev, -1); 2417 } 2418 2419 static inline void *netdev_get_ml_priv(struct net_device *dev, 2420 enum netdev_ml_priv_type type) 2421 { 2422 if (dev->ml_priv_type != type) 2423 return NULL; 2424 2425 return dev->ml_priv; 2426 } 2427 2428 static inline void netdev_set_ml_priv(struct net_device *dev, 2429 void *ml_priv, 2430 enum netdev_ml_priv_type type) 2431 { 2432 WARN(dev->ml_priv_type && dev->ml_priv_type != type, 2433 "Overwriting already set ml_priv_type (%u) with different ml_priv_type (%u)!\n", 2434 dev->ml_priv_type, type); 2435 WARN(!dev->ml_priv_type && dev->ml_priv, 2436 "Overwriting already set ml_priv and ml_priv_type is ML_PRIV_NONE!\n"); 2437 2438 dev->ml_priv = ml_priv; 2439 dev->ml_priv_type = type; 2440 } 2441 2442 /* 2443 * Net namespace inlines 2444 */ 2445 static inline 2446 struct net *dev_net(const struct net_device *dev) 2447 { 2448 return read_pnet(&dev->nd_net); 2449 } 2450 2451 static inline 2452 void dev_net_set(struct net_device *dev, struct net *net) 2453 { 2454 write_pnet(&dev->nd_net, net); 2455 } 2456 2457 /** 2458 * netdev_priv - access network device private data 2459 * @dev: network device 2460 * 2461 * Get network device private data 2462 */ 2463 static inline void *netdev_priv(const struct net_device *dev) 2464 { 2465 return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN); 2466 } 2467 2468 /* Set the sysfs physical device reference for the network logical device 2469 * if set prior to registration will cause a symlink during initialization. 2470 */ 2471 #define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev)) 2472 2473 /* Set the sysfs device type for the network logical device to allow 2474 * fine-grained identification of different network device types. For 2475 * example Ethernet, Wireless LAN, Bluetooth, WiMAX etc. 2476 */ 2477 #define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype)) 2478 2479 /* Default NAPI poll() weight 2480 * Device drivers are strongly advised to not use bigger value 2481 */ 2482 #define NAPI_POLL_WEIGHT 64 2483 2484 /** 2485 * netif_napi_add - initialize a NAPI context 2486 * @dev: network device 2487 * @napi: NAPI context 2488 * @poll: polling function 2489 * @weight: default weight 2490 * 2491 * netif_napi_add() must be used to initialize a NAPI context prior to calling 2492 * *any* of the other NAPI-related functions. 2493 */ 2494 void netif_napi_add(struct net_device *dev, struct napi_struct *napi, 2495 int (*poll)(struct napi_struct *, int), int weight); 2496 2497 /** 2498 * netif_tx_napi_add - initialize a NAPI context 2499 * @dev: network device 2500 * @napi: NAPI context 2501 * @poll: polling function 2502 * @weight: default weight 2503 * 2504 * This variant of netif_napi_add() should be used from drivers using NAPI 2505 * to exclusively poll a TX queue. 2506 * This will avoid we add it into napi_hash[], thus polluting this hash table. 2507 */ 2508 static inline void netif_tx_napi_add(struct net_device *dev, 2509 struct napi_struct *napi, 2510 int (*poll)(struct napi_struct *, int), 2511 int weight) 2512 { 2513 set_bit(NAPI_STATE_NO_BUSY_POLL, &napi->state); 2514 netif_napi_add(dev, napi, poll, weight); 2515 } 2516 2517 /** 2518 * __netif_napi_del - remove a NAPI context 2519 * @napi: NAPI context 2520 * 2521 * Warning: caller must observe RCU grace period before freeing memory 2522 * containing @napi. Drivers might want to call this helper to combine 2523 * all the needed RCU grace periods into a single one. 2524 */ 2525 void __netif_napi_del(struct napi_struct *napi); 2526 2527 /** 2528 * netif_napi_del - remove a NAPI context 2529 * @napi: NAPI context 2530 * 2531 * netif_napi_del() removes a NAPI context from the network device NAPI list 2532 */ 2533 static inline void netif_napi_del(struct napi_struct *napi) 2534 { 2535 __netif_napi_del(napi); 2536 synchronize_net(); 2537 } 2538 2539 struct packet_type { 2540 __be16 type; /* This is really htons(ether_type). */ 2541 bool ignore_outgoing; 2542 struct net_device *dev; /* NULL is wildcarded here */ 2543 int (*func) (struct sk_buff *, 2544 struct net_device *, 2545 struct packet_type *, 2546 struct net_device *); 2547 void (*list_func) (struct list_head *, 2548 struct packet_type *, 2549 struct net_device *); 2550 bool (*id_match)(struct packet_type *ptype, 2551 struct sock *sk); 2552 void *af_packet_priv; 2553 struct list_head list; 2554 }; 2555 2556 struct offload_callbacks { 2557 struct sk_buff *(*gso_segment)(struct sk_buff *skb, 2558 netdev_features_t features); 2559 struct sk_buff *(*gro_receive)(struct list_head *head, 2560 struct sk_buff *skb); 2561 int (*gro_complete)(struct sk_buff *skb, int nhoff); 2562 }; 2563 2564 struct packet_offload { 2565 __be16 type; /* This is really htons(ether_type). */ 2566 u16 priority; 2567 struct offload_callbacks callbacks; 2568 struct list_head list; 2569 }; 2570 2571 /* often modified stats are per-CPU, other are shared (netdev->stats) */ 2572 struct pcpu_sw_netstats { 2573 u64 rx_packets; 2574 u64 rx_bytes; 2575 u64 tx_packets; 2576 u64 tx_bytes; 2577 struct u64_stats_sync syncp; 2578 } __aligned(4 * sizeof(u64)); 2579 2580 struct pcpu_lstats { 2581 u64_stats_t packets; 2582 u64_stats_t bytes; 2583 struct u64_stats_sync syncp; 2584 } __aligned(2 * sizeof(u64)); 2585 2586 void dev_lstats_read(struct net_device *dev, u64 *packets, u64 *bytes); 2587 2588 static inline void dev_sw_netstats_rx_add(struct net_device *dev, unsigned int len) 2589 { 2590 struct pcpu_sw_netstats *tstats = this_cpu_ptr(dev->tstats); 2591 2592 u64_stats_update_begin(&tstats->syncp); 2593 tstats->rx_bytes += len; 2594 tstats->rx_packets++; 2595 u64_stats_update_end(&tstats->syncp); 2596 } 2597 2598 static inline void dev_sw_netstats_tx_add(struct net_device *dev, 2599 unsigned int packets, 2600 unsigned int len) 2601 { 2602 struct pcpu_sw_netstats *tstats = this_cpu_ptr(dev->tstats); 2603 2604 u64_stats_update_begin(&tstats->syncp); 2605 tstats->tx_bytes += len; 2606 tstats->tx_packets += packets; 2607 u64_stats_update_end(&tstats->syncp); 2608 } 2609 2610 static inline void dev_lstats_add(struct net_device *dev, unsigned int len) 2611 { 2612 struct pcpu_lstats *lstats = this_cpu_ptr(dev->lstats); 2613 2614 u64_stats_update_begin(&lstats->syncp); 2615 u64_stats_add(&lstats->bytes, len); 2616 u64_stats_inc(&lstats->packets); 2617 u64_stats_update_end(&lstats->syncp); 2618 } 2619 2620 #define __netdev_alloc_pcpu_stats(type, gfp) \ 2621 ({ \ 2622 typeof(type) __percpu *pcpu_stats = alloc_percpu_gfp(type, gfp);\ 2623 if (pcpu_stats) { \ 2624 int __cpu; \ 2625 for_each_possible_cpu(__cpu) { \ 2626 typeof(type) *stat; \ 2627 stat = per_cpu_ptr(pcpu_stats, __cpu); \ 2628 u64_stats_init(&stat->syncp); \ 2629 } \ 2630 } \ 2631 pcpu_stats; \ 2632 }) 2633 2634 #define netdev_alloc_pcpu_stats(type) \ 2635 __netdev_alloc_pcpu_stats(type, GFP_KERNEL) 2636 2637 #define devm_netdev_alloc_pcpu_stats(dev, type) \ 2638 ({ \ 2639 typeof(type) __percpu *pcpu_stats = devm_alloc_percpu(dev, type);\ 2640 if (pcpu_stats) { \ 2641 int __cpu; \ 2642 for_each_possible_cpu(__cpu) { \ 2643 typeof(type) *stat; \ 2644 stat = per_cpu_ptr(pcpu_stats, __cpu); \ 2645 u64_stats_init(&stat->syncp); \ 2646 } \ 2647 } \ 2648 pcpu_stats; \ 2649 }) 2650 2651 enum netdev_lag_tx_type { 2652 NETDEV_LAG_TX_TYPE_UNKNOWN, 2653 NETDEV_LAG_TX_TYPE_RANDOM, 2654 NETDEV_LAG_TX_TYPE_BROADCAST, 2655 NETDEV_LAG_TX_TYPE_ROUNDROBIN, 2656 NETDEV_LAG_TX_TYPE_ACTIVEBACKUP, 2657 NETDEV_LAG_TX_TYPE_HASH, 2658 }; 2659 2660 enum netdev_lag_hash { 2661 NETDEV_LAG_HASH_NONE, 2662 NETDEV_LAG_HASH_L2, 2663 NETDEV_LAG_HASH_L34, 2664 NETDEV_LAG_HASH_L23, 2665 NETDEV_LAG_HASH_E23, 2666 NETDEV_LAG_HASH_E34, 2667 NETDEV_LAG_HASH_VLAN_SRCMAC, 2668 NETDEV_LAG_HASH_UNKNOWN, 2669 }; 2670 2671 struct netdev_lag_upper_info { 2672 enum netdev_lag_tx_type tx_type; 2673 enum netdev_lag_hash hash_type; 2674 }; 2675 2676 struct netdev_lag_lower_state_info { 2677 u8 link_up : 1, 2678 tx_enabled : 1; 2679 }; 2680 2681 #include <linux/notifier.h> 2682 2683 /* netdevice notifier chain. Please remember to update netdev_cmd_to_name() 2684 * and the rtnetlink notification exclusion list in rtnetlink_event() when 2685 * adding new types. 2686 */ 2687 enum netdev_cmd { 2688 NETDEV_UP = 1, /* For now you can't veto a device up/down */ 2689 NETDEV_DOWN, 2690 NETDEV_REBOOT, /* Tell a protocol stack a network interface 2691 detected a hardware crash and restarted 2692 - we can use this eg to kick tcp sessions 2693 once done */ 2694 NETDEV_CHANGE, /* Notify device state change */ 2695 NETDEV_REGISTER, 2696 NETDEV_UNREGISTER, 2697 NETDEV_CHANGEMTU, /* notify after mtu change happened */ 2698 NETDEV_CHANGEADDR, /* notify after the address change */ 2699 NETDEV_PRE_CHANGEADDR, /* notify before the address change */ 2700 NETDEV_GOING_DOWN, 2701 NETDEV_CHANGENAME, 2702 NETDEV_FEAT_CHANGE, 2703 NETDEV_BONDING_FAILOVER, 2704 NETDEV_PRE_UP, 2705 NETDEV_PRE_TYPE_CHANGE, 2706 NETDEV_POST_TYPE_CHANGE, 2707 NETDEV_POST_INIT, 2708 NETDEV_RELEASE, 2709 NETDEV_NOTIFY_PEERS, 2710 NETDEV_JOIN, 2711 NETDEV_CHANGEUPPER, 2712 NETDEV_RESEND_IGMP, 2713 NETDEV_PRECHANGEMTU, /* notify before mtu change happened */ 2714 NETDEV_CHANGEINFODATA, 2715 NETDEV_BONDING_INFO, 2716 NETDEV_PRECHANGEUPPER, 2717 NETDEV_CHANGELOWERSTATE, 2718 NETDEV_UDP_TUNNEL_PUSH_INFO, 2719 NETDEV_UDP_TUNNEL_DROP_INFO, 2720 NETDEV_CHANGE_TX_QUEUE_LEN, 2721 NETDEV_CVLAN_FILTER_PUSH_INFO, 2722 NETDEV_CVLAN_FILTER_DROP_INFO, 2723 NETDEV_SVLAN_FILTER_PUSH_INFO, 2724 NETDEV_SVLAN_FILTER_DROP_INFO, 2725 }; 2726 const char *netdev_cmd_to_name(enum netdev_cmd cmd); 2727 2728 int register_netdevice_notifier(struct notifier_block *nb); 2729 int unregister_netdevice_notifier(struct notifier_block *nb); 2730 int register_netdevice_notifier_net(struct net *net, struct notifier_block *nb); 2731 int unregister_netdevice_notifier_net(struct net *net, 2732 struct notifier_block *nb); 2733 int register_netdevice_notifier_dev_net(struct net_device *dev, 2734 struct notifier_block *nb, 2735 struct netdev_net_notifier *nn); 2736 int unregister_netdevice_notifier_dev_net(struct net_device *dev, 2737 struct notifier_block *nb, 2738 struct netdev_net_notifier *nn); 2739 2740 struct netdev_notifier_info { 2741 struct net_device *dev; 2742 struct netlink_ext_ack *extack; 2743 }; 2744 2745 struct netdev_notifier_info_ext { 2746 struct netdev_notifier_info info; /* must be first */ 2747 union { 2748 u32 mtu; 2749 } ext; 2750 }; 2751 2752 struct netdev_notifier_change_info { 2753 struct netdev_notifier_info info; /* must be first */ 2754 unsigned int flags_changed; 2755 }; 2756 2757 struct netdev_notifier_changeupper_info { 2758 struct netdev_notifier_info info; /* must be first */ 2759 struct net_device *upper_dev; /* new upper dev */ 2760 bool master; /* is upper dev master */ 2761 bool linking; /* is the notification for link or unlink */ 2762 void *upper_info; /* upper dev info */ 2763 }; 2764 2765 struct netdev_notifier_changelowerstate_info { 2766 struct netdev_notifier_info info; /* must be first */ 2767 void *lower_state_info; /* is lower dev state */ 2768 }; 2769 2770 struct netdev_notifier_pre_changeaddr_info { 2771 struct netdev_notifier_info info; /* must be first */ 2772 const unsigned char *dev_addr; 2773 }; 2774 2775 static inline void netdev_notifier_info_init(struct netdev_notifier_info *info, 2776 struct net_device *dev) 2777 { 2778 info->dev = dev; 2779 info->extack = NULL; 2780 } 2781 2782 static inline struct net_device * 2783 netdev_notifier_info_to_dev(const struct netdev_notifier_info *info) 2784 { 2785 return info->dev; 2786 } 2787 2788 static inline struct netlink_ext_ack * 2789 netdev_notifier_info_to_extack(const struct netdev_notifier_info *info) 2790 { 2791 return info->extack; 2792 } 2793 2794 int call_netdevice_notifiers(unsigned long val, struct net_device *dev); 2795 2796 2797 extern rwlock_t dev_base_lock; /* Device list lock */ 2798 2799 #define for_each_netdev(net, d) \ 2800 list_for_each_entry(d, &(net)->dev_base_head, dev_list) 2801 #define for_each_netdev_reverse(net, d) \ 2802 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list) 2803 #define for_each_netdev_rcu(net, d) \ 2804 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list) 2805 #define for_each_netdev_safe(net, d, n) \ 2806 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list) 2807 #define for_each_netdev_continue(net, d) \ 2808 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list) 2809 #define for_each_netdev_continue_reverse(net, d) \ 2810 list_for_each_entry_continue_reverse(d, &(net)->dev_base_head, \ 2811 dev_list) 2812 #define for_each_netdev_continue_rcu(net, d) \ 2813 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list) 2814 #define for_each_netdev_in_bond_rcu(bond, slave) \ 2815 for_each_netdev_rcu(&init_net, slave) \ 2816 if (netdev_master_upper_dev_get_rcu(slave) == (bond)) 2817 #define net_device_entry(lh) list_entry(lh, struct net_device, dev_list) 2818 2819 static inline struct net_device *next_net_device(struct net_device *dev) 2820 { 2821 struct list_head *lh; 2822 struct net *net; 2823 2824 net = dev_net(dev); 2825 lh = dev->dev_list.next; 2826 return lh == &net->dev_base_head ? NULL : net_device_entry(lh); 2827 } 2828 2829 static inline struct net_device *next_net_device_rcu(struct net_device *dev) 2830 { 2831 struct list_head *lh; 2832 struct net *net; 2833 2834 net = dev_net(dev); 2835 lh = rcu_dereference(list_next_rcu(&dev->dev_list)); 2836 return lh == &net->dev_base_head ? NULL : net_device_entry(lh); 2837 } 2838 2839 static inline struct net_device *first_net_device(struct net *net) 2840 { 2841 return list_empty(&net->dev_base_head) ? NULL : 2842 net_device_entry(net->dev_base_head.next); 2843 } 2844 2845 static inline struct net_device *first_net_device_rcu(struct net *net) 2846 { 2847 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head)); 2848 2849 return lh == &net->dev_base_head ? NULL : net_device_entry(lh); 2850 } 2851 2852 int netdev_boot_setup_check(struct net_device *dev); 2853 struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type, 2854 const char *hwaddr); 2855 struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type); 2856 void dev_add_pack(struct packet_type *pt); 2857 void dev_remove_pack(struct packet_type *pt); 2858 void __dev_remove_pack(struct packet_type *pt); 2859 void dev_add_offload(struct packet_offload *po); 2860 void dev_remove_offload(struct packet_offload *po); 2861 2862 int dev_get_iflink(const struct net_device *dev); 2863 int dev_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb); 2864 int dev_fill_forward_path(const struct net_device *dev, const u8 *daddr, 2865 struct net_device_path_stack *stack); 2866 struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags, 2867 unsigned short mask); 2868 struct net_device *dev_get_by_name(struct net *net, const char *name); 2869 struct net_device *dev_get_by_name_rcu(struct net *net, const char *name); 2870 struct net_device *__dev_get_by_name(struct net *net, const char *name); 2871 bool netdev_name_in_use(struct net *net, const char *name); 2872 int dev_alloc_name(struct net_device *dev, const char *name); 2873 int dev_open(struct net_device *dev, struct netlink_ext_ack *extack); 2874 void dev_close(struct net_device *dev); 2875 void dev_close_many(struct list_head *head, bool unlink); 2876 void dev_disable_lro(struct net_device *dev); 2877 int dev_loopback_xmit(struct net *net, struct sock *sk, struct sk_buff *newskb); 2878 u16 dev_pick_tx_zero(struct net_device *dev, struct sk_buff *skb, 2879 struct net_device *sb_dev); 2880 u16 dev_pick_tx_cpu_id(struct net_device *dev, struct sk_buff *skb, 2881 struct net_device *sb_dev); 2882 2883 int dev_queue_xmit(struct sk_buff *skb); 2884 int dev_queue_xmit_accel(struct sk_buff *skb, struct net_device *sb_dev); 2885 int __dev_direct_xmit(struct sk_buff *skb, u16 queue_id); 2886 2887 static inline int dev_direct_xmit(struct sk_buff *skb, u16 queue_id) 2888 { 2889 int ret; 2890 2891 ret = __dev_direct_xmit(skb, queue_id); 2892 if (!dev_xmit_complete(ret)) 2893 kfree_skb(skb); 2894 return ret; 2895 } 2896 2897 int register_netdevice(struct net_device *dev); 2898 void unregister_netdevice_queue(struct net_device *dev, struct list_head *head); 2899 void unregister_netdevice_many(struct list_head *head); 2900 static inline void unregister_netdevice(struct net_device *dev) 2901 { 2902 unregister_netdevice_queue(dev, NULL); 2903 } 2904 2905 int netdev_refcnt_read(const struct net_device *dev); 2906 void free_netdev(struct net_device *dev); 2907 void netdev_freemem(struct net_device *dev); 2908 int init_dummy_netdev(struct net_device *dev); 2909 2910 struct net_device *netdev_get_xmit_slave(struct net_device *dev, 2911 struct sk_buff *skb, 2912 bool all_slaves); 2913 struct net_device *netdev_sk_get_lowest_dev(struct net_device *dev, 2914 struct sock *sk); 2915 struct net_device *dev_get_by_index(struct net *net, int ifindex); 2916 struct net_device *__dev_get_by_index(struct net *net, int ifindex); 2917 struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex); 2918 struct net_device *dev_get_by_napi_id(unsigned int napi_id); 2919 int netdev_get_name(struct net *net, char *name, int ifindex); 2920 int dev_restart(struct net_device *dev); 2921 2922 2923 static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev, 2924 unsigned short type, 2925 const void *daddr, const void *saddr, 2926 unsigned int len) 2927 { 2928 if (!dev->header_ops || !dev->header_ops->create) 2929 return 0; 2930 2931 return dev->header_ops->create(skb, dev, type, daddr, saddr, len); 2932 } 2933 2934 static inline int dev_parse_header(const struct sk_buff *skb, 2935 unsigned char *haddr) 2936 { 2937 const struct net_device *dev = skb->dev; 2938 2939 if (!dev->header_ops || !dev->header_ops->parse) 2940 return 0; 2941 return dev->header_ops->parse(skb, haddr); 2942 } 2943 2944 static inline __be16 dev_parse_header_protocol(const struct sk_buff *skb) 2945 { 2946 const struct net_device *dev = skb->dev; 2947 2948 if (!dev->header_ops || !dev->header_ops->parse_protocol) 2949 return 0; 2950 return dev->header_ops->parse_protocol(skb); 2951 } 2952 2953 /* ll_header must have at least hard_header_len allocated */ 2954 static inline bool dev_validate_header(const struct net_device *dev, 2955 char *ll_header, int len) 2956 { 2957 if (likely(len >= dev->hard_header_len)) 2958 return true; 2959 if (len < dev->min_header_len) 2960 return false; 2961 2962 if (capable(CAP_SYS_RAWIO)) { 2963 memset(ll_header + len, 0, dev->hard_header_len - len); 2964 return true; 2965 } 2966 2967 if (dev->header_ops && dev->header_ops->validate) 2968 return dev->header_ops->validate(ll_header, len); 2969 2970 return false; 2971 } 2972 2973 static inline bool dev_has_header(const struct net_device *dev) 2974 { 2975 return dev->header_ops && dev->header_ops->create; 2976 } 2977 2978 #ifdef CONFIG_NET_FLOW_LIMIT 2979 #define FLOW_LIMIT_HISTORY (1 << 7) /* must be ^2 and !overflow buckets */ 2980 struct sd_flow_limit { 2981 u64 count; 2982 unsigned int num_buckets; 2983 unsigned int history_head; 2984 u16 history[FLOW_LIMIT_HISTORY]; 2985 u8 buckets[]; 2986 }; 2987 2988 extern int netdev_flow_limit_table_len; 2989 #endif /* CONFIG_NET_FLOW_LIMIT */ 2990 2991 /* 2992 * Incoming packets are placed on per-CPU queues 2993 */ 2994 struct softnet_data { 2995 struct list_head poll_list; 2996 struct sk_buff_head process_queue; 2997 2998 /* stats */ 2999 unsigned int processed; 3000 unsigned int time_squeeze; 3001 unsigned int received_rps; 3002 #ifdef CONFIG_RPS 3003 struct softnet_data *rps_ipi_list; 3004 #endif 3005 #ifdef CONFIG_NET_FLOW_LIMIT 3006 struct sd_flow_limit __rcu *flow_limit; 3007 #endif 3008 struct Qdisc *output_queue; 3009 struct Qdisc **output_queue_tailp; 3010 struct sk_buff *completion_queue; 3011 #ifdef CONFIG_XFRM_OFFLOAD 3012 struct sk_buff_head xfrm_backlog; 3013 #endif 3014 /* written and read only by owning cpu: */ 3015 struct { 3016 u16 recursion; 3017 u8 more; 3018 } xmit; 3019 #ifdef CONFIG_RPS 3020 /* input_queue_head should be written by cpu owning this struct, 3021 * and only read by other cpus. Worth using a cache line. 3022 */ 3023 unsigned int input_queue_head ____cacheline_aligned_in_smp; 3024 3025 /* Elements below can be accessed between CPUs for RPS/RFS */ 3026 call_single_data_t csd ____cacheline_aligned_in_smp; 3027 struct softnet_data *rps_ipi_next; 3028 unsigned int cpu; 3029 unsigned int input_queue_tail; 3030 #endif 3031 unsigned int dropped; 3032 struct sk_buff_head input_pkt_queue; 3033 struct napi_struct backlog; 3034 3035 }; 3036 3037 static inline void input_queue_head_incr(struct softnet_data *sd) 3038 { 3039 #ifdef CONFIG_RPS 3040 sd->input_queue_head++; 3041 #endif 3042 } 3043 3044 static inline void input_queue_tail_incr_save(struct softnet_data *sd, 3045 unsigned int *qtail) 3046 { 3047 #ifdef CONFIG_RPS 3048 *qtail = ++sd->input_queue_tail; 3049 #endif 3050 } 3051 3052 DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data); 3053 3054 static inline int dev_recursion_level(void) 3055 { 3056 return this_cpu_read(softnet_data.xmit.recursion); 3057 } 3058 3059 #define XMIT_RECURSION_LIMIT 8 3060 static inline bool dev_xmit_recursion(void) 3061 { 3062 return unlikely(__this_cpu_read(softnet_data.xmit.recursion) > 3063 XMIT_RECURSION_LIMIT); 3064 } 3065 3066 static inline void dev_xmit_recursion_inc(void) 3067 { 3068 __this_cpu_inc(softnet_data.xmit.recursion); 3069 } 3070 3071 static inline void dev_xmit_recursion_dec(void) 3072 { 3073 __this_cpu_dec(softnet_data.xmit.recursion); 3074 } 3075 3076 void __netif_schedule(struct Qdisc *q); 3077 void netif_schedule_queue(struct netdev_queue *txq); 3078 3079 static inline void netif_tx_schedule_all(struct net_device *dev) 3080 { 3081 unsigned int i; 3082 3083 for (i = 0; i < dev->num_tx_queues; i++) 3084 netif_schedule_queue(netdev_get_tx_queue(dev, i)); 3085 } 3086 3087 static __always_inline void netif_tx_start_queue(struct netdev_queue *dev_queue) 3088 { 3089 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state); 3090 } 3091 3092 /** 3093 * netif_start_queue - allow transmit 3094 * @dev: network device 3095 * 3096 * Allow upper layers to call the device hard_start_xmit routine. 3097 */ 3098 static inline void netif_start_queue(struct net_device *dev) 3099 { 3100 netif_tx_start_queue(netdev_get_tx_queue(dev, 0)); 3101 } 3102 3103 static inline void netif_tx_start_all_queues(struct net_device *dev) 3104 { 3105 unsigned int i; 3106 3107 for (i = 0; i < dev->num_tx_queues; i++) { 3108 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 3109 netif_tx_start_queue(txq); 3110 } 3111 } 3112 3113 void netif_tx_wake_queue(struct netdev_queue *dev_queue); 3114 3115 /** 3116 * netif_wake_queue - restart transmit 3117 * @dev: network device 3118 * 3119 * Allow upper layers to call the device hard_start_xmit routine. 3120 * Used for flow control when transmit resources are available. 3121 */ 3122 static inline void netif_wake_queue(struct net_device *dev) 3123 { 3124 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0)); 3125 } 3126 3127 static inline void netif_tx_wake_all_queues(struct net_device *dev) 3128 { 3129 unsigned int i; 3130 3131 for (i = 0; i < dev->num_tx_queues; i++) { 3132 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 3133 netif_tx_wake_queue(txq); 3134 } 3135 } 3136 3137 static __always_inline void netif_tx_stop_queue(struct netdev_queue *dev_queue) 3138 { 3139 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state); 3140 } 3141 3142 /** 3143 * netif_stop_queue - stop transmitted packets 3144 * @dev: network device 3145 * 3146 * Stop upper layers calling the device hard_start_xmit routine. 3147 * Used for flow control when transmit resources are unavailable. 3148 */ 3149 static inline void netif_stop_queue(struct net_device *dev) 3150 { 3151 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0)); 3152 } 3153 3154 void netif_tx_stop_all_queues(struct net_device *dev); 3155 3156 static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue) 3157 { 3158 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state); 3159 } 3160 3161 /** 3162 * netif_queue_stopped - test if transmit queue is flowblocked 3163 * @dev: network device 3164 * 3165 * Test if transmit queue on device is currently unable to send. 3166 */ 3167 static inline bool netif_queue_stopped(const struct net_device *dev) 3168 { 3169 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0)); 3170 } 3171 3172 static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue) 3173 { 3174 return dev_queue->state & QUEUE_STATE_ANY_XOFF; 3175 } 3176 3177 static inline bool 3178 netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue) 3179 { 3180 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN; 3181 } 3182 3183 static inline bool 3184 netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue) 3185 { 3186 return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN; 3187 } 3188 3189 /** 3190 * netdev_queue_set_dql_min_limit - set dql minimum limit 3191 * @dev_queue: pointer to transmit queue 3192 * @min_limit: dql minimum limit 3193 * 3194 * Forces xmit_more() to return true until the minimum threshold 3195 * defined by @min_limit is reached (or until the tx queue is 3196 * empty). Warning: to be use with care, misuse will impact the 3197 * latency. 3198 */ 3199 static inline void netdev_queue_set_dql_min_limit(struct netdev_queue *dev_queue, 3200 unsigned int min_limit) 3201 { 3202 #ifdef CONFIG_BQL 3203 dev_queue->dql.min_limit = min_limit; 3204 #endif 3205 } 3206 3207 /** 3208 * netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write 3209 * @dev_queue: pointer to transmit queue 3210 * 3211 * BQL enabled drivers might use this helper in their ndo_start_xmit(), 3212 * to give appropriate hint to the CPU. 3213 */ 3214 static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue) 3215 { 3216 #ifdef CONFIG_BQL 3217 prefetchw(&dev_queue->dql.num_queued); 3218 #endif 3219 } 3220 3221 /** 3222 * netdev_txq_bql_complete_prefetchw - prefetch bql data for write 3223 * @dev_queue: pointer to transmit queue 3224 * 3225 * BQL enabled drivers might use this helper in their TX completion path, 3226 * to give appropriate hint to the CPU. 3227 */ 3228 static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue) 3229 { 3230 #ifdef CONFIG_BQL 3231 prefetchw(&dev_queue->dql.limit); 3232 #endif 3233 } 3234 3235 static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue, 3236 unsigned int bytes) 3237 { 3238 #ifdef CONFIG_BQL 3239 dql_queued(&dev_queue->dql, bytes); 3240 3241 if (likely(dql_avail(&dev_queue->dql) >= 0)) 3242 return; 3243 3244 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state); 3245 3246 /* 3247 * The XOFF flag must be set before checking the dql_avail below, 3248 * because in netdev_tx_completed_queue we update the dql_completed 3249 * before checking the XOFF flag. 3250 */ 3251 smp_mb(); 3252 3253 /* check again in case another CPU has just made room avail */ 3254 if (unlikely(dql_avail(&dev_queue->dql) >= 0)) 3255 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state); 3256 #endif 3257 } 3258 3259 /* Variant of netdev_tx_sent_queue() for drivers that are aware 3260 * that they should not test BQL status themselves. 3261 * We do want to change __QUEUE_STATE_STACK_XOFF only for the last 3262 * skb of a batch. 3263 * Returns true if the doorbell must be used to kick the NIC. 3264 */ 3265 static inline bool __netdev_tx_sent_queue(struct netdev_queue *dev_queue, 3266 unsigned int bytes, 3267 bool xmit_more) 3268 { 3269 if (xmit_more) { 3270 #ifdef CONFIG_BQL 3271 dql_queued(&dev_queue->dql, bytes); 3272 #endif 3273 return netif_tx_queue_stopped(dev_queue); 3274 } 3275 netdev_tx_sent_queue(dev_queue, bytes); 3276 return true; 3277 } 3278 3279 /** 3280 * netdev_sent_queue - report the number of bytes queued to hardware 3281 * @dev: network device 3282 * @bytes: number of bytes queued to the hardware device queue 3283 * 3284 * Report the number of bytes queued for sending/completion to the network 3285 * device hardware queue. @bytes should be a good approximation and should 3286 * exactly match netdev_completed_queue() @bytes 3287 */ 3288 static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes) 3289 { 3290 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes); 3291 } 3292 3293 static inline bool __netdev_sent_queue(struct net_device *dev, 3294 unsigned int bytes, 3295 bool xmit_more) 3296 { 3297 return __netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes, 3298 xmit_more); 3299 } 3300 3301 static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue, 3302 unsigned int pkts, unsigned int bytes) 3303 { 3304 #ifdef CONFIG_BQL 3305 if (unlikely(!bytes)) 3306 return; 3307 3308 dql_completed(&dev_queue->dql, bytes); 3309 3310 /* 3311 * Without the memory barrier there is a small possiblity that 3312 * netdev_tx_sent_queue will miss the update and cause the queue to 3313 * be stopped forever 3314 */ 3315 smp_mb(); 3316 3317 if (unlikely(dql_avail(&dev_queue->dql) < 0)) 3318 return; 3319 3320 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state)) 3321 netif_schedule_queue(dev_queue); 3322 #endif 3323 } 3324 3325 /** 3326 * netdev_completed_queue - report bytes and packets completed by device 3327 * @dev: network device 3328 * @pkts: actual number of packets sent over the medium 3329 * @bytes: actual number of bytes sent over the medium 3330 * 3331 * Report the number of bytes and packets transmitted by the network device 3332 * hardware queue over the physical medium, @bytes must exactly match the 3333 * @bytes amount passed to netdev_sent_queue() 3334 */ 3335 static inline void netdev_completed_queue(struct net_device *dev, 3336 unsigned int pkts, unsigned int bytes) 3337 { 3338 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes); 3339 } 3340 3341 static inline void netdev_tx_reset_queue(struct netdev_queue *q) 3342 { 3343 #ifdef CONFIG_BQL 3344 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state); 3345 dql_reset(&q->dql); 3346 #endif 3347 } 3348 3349 /** 3350 * netdev_reset_queue - reset the packets and bytes count of a network device 3351 * @dev_queue: network device 3352 * 3353 * Reset the bytes and packet count of a network device and clear the 3354 * software flow control OFF bit for this network device 3355 */ 3356 static inline void netdev_reset_queue(struct net_device *dev_queue) 3357 { 3358 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0)); 3359 } 3360 3361 /** 3362 * netdev_cap_txqueue - check if selected tx queue exceeds device queues 3363 * @dev: network device 3364 * @queue_index: given tx queue index 3365 * 3366 * Returns 0 if given tx queue index >= number of device tx queues, 3367 * otherwise returns the originally passed tx queue index. 3368 */ 3369 static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index) 3370 { 3371 if (unlikely(queue_index >= dev->real_num_tx_queues)) { 3372 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n", 3373 dev->name, queue_index, 3374 dev->real_num_tx_queues); 3375 return 0; 3376 } 3377 3378 return queue_index; 3379 } 3380 3381 /** 3382 * netif_running - test if up 3383 * @dev: network device 3384 * 3385 * Test if the device has been brought up. 3386 */ 3387 static inline bool netif_running(const struct net_device *dev) 3388 { 3389 return test_bit(__LINK_STATE_START, &dev->state); 3390 } 3391 3392 /* 3393 * Routines to manage the subqueues on a device. We only need start, 3394 * stop, and a check if it's stopped. All other device management is 3395 * done at the overall netdevice level. 3396 * Also test the device if we're multiqueue. 3397 */ 3398 3399 /** 3400 * netif_start_subqueue - allow sending packets on subqueue 3401 * @dev: network device 3402 * @queue_index: sub queue index 3403 * 3404 * Start individual transmit queue of a device with multiple transmit queues. 3405 */ 3406 static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index) 3407 { 3408 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 3409 3410 netif_tx_start_queue(txq); 3411 } 3412 3413 /** 3414 * netif_stop_subqueue - stop sending packets on subqueue 3415 * @dev: network device 3416 * @queue_index: sub queue index 3417 * 3418 * Stop individual transmit queue of a device with multiple transmit queues. 3419 */ 3420 static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index) 3421 { 3422 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 3423 netif_tx_stop_queue(txq); 3424 } 3425 3426 /** 3427 * __netif_subqueue_stopped - test status of subqueue 3428 * @dev: network device 3429 * @queue_index: sub queue index 3430 * 3431 * Check individual transmit queue of a device with multiple transmit queues. 3432 */ 3433 static inline bool __netif_subqueue_stopped(const struct net_device *dev, 3434 u16 queue_index) 3435 { 3436 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 3437 3438 return netif_tx_queue_stopped(txq); 3439 } 3440 3441 /** 3442 * netif_subqueue_stopped - test status of subqueue 3443 * @dev: network device 3444 * @skb: sub queue buffer pointer 3445 * 3446 * Check individual transmit queue of a device with multiple transmit queues. 3447 */ 3448 static inline bool netif_subqueue_stopped(const struct net_device *dev, 3449 struct sk_buff *skb) 3450 { 3451 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb)); 3452 } 3453 3454 /** 3455 * netif_wake_subqueue - allow sending packets on subqueue 3456 * @dev: network device 3457 * @queue_index: sub queue index 3458 * 3459 * Resume individual transmit queue of a device with multiple transmit queues. 3460 */ 3461 static inline void netif_wake_subqueue(struct net_device *dev, u16 queue_index) 3462 { 3463 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 3464 3465 netif_tx_wake_queue(txq); 3466 } 3467 3468 #ifdef CONFIG_XPS 3469 int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask, 3470 u16 index); 3471 int __netif_set_xps_queue(struct net_device *dev, const unsigned long *mask, 3472 u16 index, enum xps_map_type type); 3473 3474 /** 3475 * netif_attr_test_mask - Test a CPU or Rx queue set in a mask 3476 * @j: CPU/Rx queue index 3477 * @mask: bitmask of all cpus/rx queues 3478 * @nr_bits: number of bits in the bitmask 3479 * 3480 * Test if a CPU or Rx queue index is set in a mask of all CPU/Rx queues. 3481 */ 3482 static inline bool netif_attr_test_mask(unsigned long j, 3483 const unsigned long *mask, 3484 unsigned int nr_bits) 3485 { 3486 cpu_max_bits_warn(j, nr_bits); 3487 return test_bit(j, mask); 3488 } 3489 3490 /** 3491 * netif_attr_test_online - Test for online CPU/Rx queue 3492 * @j: CPU/Rx queue index 3493 * @online_mask: bitmask for CPUs/Rx queues that are online 3494 * @nr_bits: number of bits in the bitmask 3495 * 3496 * Returns true if a CPU/Rx queue is online. 3497 */ 3498 static inline bool netif_attr_test_online(unsigned long j, 3499 const unsigned long *online_mask, 3500 unsigned int nr_bits) 3501 { 3502 cpu_max_bits_warn(j, nr_bits); 3503 3504 if (online_mask) 3505 return test_bit(j, online_mask); 3506 3507 return (j < nr_bits); 3508 } 3509 3510 /** 3511 * netif_attrmask_next - get the next CPU/Rx queue in a cpu/Rx queues mask 3512 * @n: CPU/Rx queue index 3513 * @srcp: the cpumask/Rx queue mask pointer 3514 * @nr_bits: number of bits in the bitmask 3515 * 3516 * Returns >= nr_bits if no further CPUs/Rx queues set. 3517 */ 3518 static inline unsigned int netif_attrmask_next(int n, const unsigned long *srcp, 3519 unsigned int nr_bits) 3520 { 3521 /* -1 is a legal arg here. */ 3522 if (n != -1) 3523 cpu_max_bits_warn(n, nr_bits); 3524 3525 if (srcp) 3526 return find_next_bit(srcp, nr_bits, n + 1); 3527 3528 return n + 1; 3529 } 3530 3531 /** 3532 * netif_attrmask_next_and - get the next CPU/Rx queue in \*src1p & \*src2p 3533 * @n: CPU/Rx queue index 3534 * @src1p: the first CPUs/Rx queues mask pointer 3535 * @src2p: the second CPUs/Rx queues mask pointer 3536 * @nr_bits: number of bits in the bitmask 3537 * 3538 * Returns >= nr_bits if no further CPUs/Rx queues set in both. 3539 */ 3540 static inline int netif_attrmask_next_and(int n, const unsigned long *src1p, 3541 const unsigned long *src2p, 3542 unsigned int nr_bits) 3543 { 3544 /* -1 is a legal arg here. */ 3545 if (n != -1) 3546 cpu_max_bits_warn(n, nr_bits); 3547 3548 if (src1p && src2p) 3549 return find_next_and_bit(src1p, src2p, nr_bits, n + 1); 3550 else if (src1p) 3551 return find_next_bit(src1p, nr_bits, n + 1); 3552 else if (src2p) 3553 return find_next_bit(src2p, nr_bits, n + 1); 3554 3555 return n + 1; 3556 } 3557 #else 3558 static inline int netif_set_xps_queue(struct net_device *dev, 3559 const struct cpumask *mask, 3560 u16 index) 3561 { 3562 return 0; 3563 } 3564 3565 static inline int __netif_set_xps_queue(struct net_device *dev, 3566 const unsigned long *mask, 3567 u16 index, enum xps_map_type type) 3568 { 3569 return 0; 3570 } 3571 #endif 3572 3573 /** 3574 * netif_is_multiqueue - test if device has multiple transmit queues 3575 * @dev: network device 3576 * 3577 * Check if device has multiple transmit queues 3578 */ 3579 static inline bool netif_is_multiqueue(const struct net_device *dev) 3580 { 3581 return dev->num_tx_queues > 1; 3582 } 3583 3584 int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq); 3585 3586 #ifdef CONFIG_SYSFS 3587 int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq); 3588 #else 3589 static inline int netif_set_real_num_rx_queues(struct net_device *dev, 3590 unsigned int rxqs) 3591 { 3592 dev->real_num_rx_queues = rxqs; 3593 return 0; 3594 } 3595 #endif 3596 int netif_set_real_num_queues(struct net_device *dev, 3597 unsigned int txq, unsigned int rxq); 3598 3599 static inline struct netdev_rx_queue * 3600 __netif_get_rx_queue(struct net_device *dev, unsigned int rxq) 3601 { 3602 return dev->_rx + rxq; 3603 } 3604 3605 #ifdef CONFIG_SYSFS 3606 static inline unsigned int get_netdev_rx_queue_index( 3607 struct netdev_rx_queue *queue) 3608 { 3609 struct net_device *dev = queue->dev; 3610 int index = queue - dev->_rx; 3611 3612 BUG_ON(index >= dev->num_rx_queues); 3613 return index; 3614 } 3615 #endif 3616 3617 #define DEFAULT_MAX_NUM_RSS_QUEUES (8) 3618 int netif_get_num_default_rss_queues(void); 3619 3620 enum skb_free_reason { 3621 SKB_REASON_CONSUMED, 3622 SKB_REASON_DROPPED, 3623 }; 3624 3625 void __dev_kfree_skb_irq(struct sk_buff *skb, enum skb_free_reason reason); 3626 void __dev_kfree_skb_any(struct sk_buff *skb, enum skb_free_reason reason); 3627 3628 /* 3629 * It is not allowed to call kfree_skb() or consume_skb() from hardware 3630 * interrupt context or with hardware interrupts being disabled. 3631 * (in_hardirq() || irqs_disabled()) 3632 * 3633 * We provide four helpers that can be used in following contexts : 3634 * 3635 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context, 3636 * replacing kfree_skb(skb) 3637 * 3638 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context. 3639 * Typically used in place of consume_skb(skb) in TX completion path 3640 * 3641 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context, 3642 * replacing kfree_skb(skb) 3643 * 3644 * dev_consume_skb_any(skb) when caller doesn't know its current irq context, 3645 * and consumed a packet. Used in place of consume_skb(skb) 3646 */ 3647 static inline void dev_kfree_skb_irq(struct sk_buff *skb) 3648 { 3649 __dev_kfree_skb_irq(skb, SKB_REASON_DROPPED); 3650 } 3651 3652 static inline void dev_consume_skb_irq(struct sk_buff *skb) 3653 { 3654 __dev_kfree_skb_irq(skb, SKB_REASON_CONSUMED); 3655 } 3656 3657 static inline void dev_kfree_skb_any(struct sk_buff *skb) 3658 { 3659 __dev_kfree_skb_any(skb, SKB_REASON_DROPPED); 3660 } 3661 3662 static inline void dev_consume_skb_any(struct sk_buff *skb) 3663 { 3664 __dev_kfree_skb_any(skb, SKB_REASON_CONSUMED); 3665 } 3666 3667 u32 bpf_prog_run_generic_xdp(struct sk_buff *skb, struct xdp_buff *xdp, 3668 struct bpf_prog *xdp_prog); 3669 void generic_xdp_tx(struct sk_buff *skb, struct bpf_prog *xdp_prog); 3670 int do_xdp_generic(struct bpf_prog *xdp_prog, struct sk_buff *skb); 3671 int netif_rx(struct sk_buff *skb); 3672 int netif_rx_ni(struct sk_buff *skb); 3673 int netif_rx_any_context(struct sk_buff *skb); 3674 int netif_receive_skb(struct sk_buff *skb); 3675 int netif_receive_skb_core(struct sk_buff *skb); 3676 void netif_receive_skb_list_internal(struct list_head *head); 3677 void netif_receive_skb_list(struct list_head *head); 3678 gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb); 3679 void napi_gro_flush(struct napi_struct *napi, bool flush_old); 3680 struct sk_buff *napi_get_frags(struct napi_struct *napi); 3681 gro_result_t napi_gro_frags(struct napi_struct *napi); 3682 struct packet_offload *gro_find_receive_by_type(__be16 type); 3683 struct packet_offload *gro_find_complete_by_type(__be16 type); 3684 3685 static inline void napi_free_frags(struct napi_struct *napi) 3686 { 3687 kfree_skb(napi->skb); 3688 napi->skb = NULL; 3689 } 3690 3691 bool netdev_is_rx_handler_busy(struct net_device *dev); 3692 int netdev_rx_handler_register(struct net_device *dev, 3693 rx_handler_func_t *rx_handler, 3694 void *rx_handler_data); 3695 void netdev_rx_handler_unregister(struct net_device *dev); 3696 3697 bool dev_valid_name(const char *name); 3698 static inline bool is_socket_ioctl_cmd(unsigned int cmd) 3699 { 3700 return _IOC_TYPE(cmd) == SOCK_IOC_TYPE; 3701 } 3702 int get_user_ifreq(struct ifreq *ifr, void __user **ifrdata, void __user *arg); 3703 int put_user_ifreq(struct ifreq *ifr, void __user *arg); 3704 int dev_ioctl(struct net *net, unsigned int cmd, struct ifreq *ifr, 3705 void __user *data, bool *need_copyout); 3706 int dev_ifconf(struct net *net, struct ifconf __user *ifc); 3707 int dev_ethtool(struct net *net, struct ifreq *ifr, void __user *userdata); 3708 unsigned int dev_get_flags(const struct net_device *); 3709 int __dev_change_flags(struct net_device *dev, unsigned int flags, 3710 struct netlink_ext_ack *extack); 3711 int dev_change_flags(struct net_device *dev, unsigned int flags, 3712 struct netlink_ext_ack *extack); 3713 void __dev_notify_flags(struct net_device *, unsigned int old_flags, 3714 unsigned int gchanges); 3715 int dev_change_name(struct net_device *, const char *); 3716 int dev_set_alias(struct net_device *, const char *, size_t); 3717 int dev_get_alias(const struct net_device *, char *, size_t); 3718 int __dev_change_net_namespace(struct net_device *dev, struct net *net, 3719 const char *pat, int new_ifindex); 3720 static inline 3721 int dev_change_net_namespace(struct net_device *dev, struct net *net, 3722 const char *pat) 3723 { 3724 return __dev_change_net_namespace(dev, net, pat, 0); 3725 } 3726 int __dev_set_mtu(struct net_device *, int); 3727 int dev_validate_mtu(struct net_device *dev, int mtu, 3728 struct netlink_ext_ack *extack); 3729 int dev_set_mtu_ext(struct net_device *dev, int mtu, 3730 struct netlink_ext_ack *extack); 3731 int dev_set_mtu(struct net_device *, int); 3732 int dev_change_tx_queue_len(struct net_device *, unsigned long); 3733 void dev_set_group(struct net_device *, int); 3734 int dev_pre_changeaddr_notify(struct net_device *dev, const char *addr, 3735 struct netlink_ext_ack *extack); 3736 int dev_set_mac_address(struct net_device *dev, struct sockaddr *sa, 3737 struct netlink_ext_ack *extack); 3738 int dev_set_mac_address_user(struct net_device *dev, struct sockaddr *sa, 3739 struct netlink_ext_ack *extack); 3740 int dev_get_mac_address(struct sockaddr *sa, struct net *net, char *dev_name); 3741 int dev_change_carrier(struct net_device *, bool new_carrier); 3742 int dev_get_phys_port_id(struct net_device *dev, 3743 struct netdev_phys_item_id *ppid); 3744 int dev_get_phys_port_name(struct net_device *dev, 3745 char *name, size_t len); 3746 int dev_get_port_parent_id(struct net_device *dev, 3747 struct netdev_phys_item_id *ppid, bool recurse); 3748 bool netdev_port_same_parent_id(struct net_device *a, struct net_device *b); 3749 int dev_change_proto_down(struct net_device *dev, bool proto_down); 3750 void dev_change_proto_down_reason(struct net_device *dev, unsigned long mask, 3751 u32 value); 3752 struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev, bool *again); 3753 struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev, 3754 struct netdev_queue *txq, int *ret); 3755 3756 typedef int (*bpf_op_t)(struct net_device *dev, struct netdev_bpf *bpf); 3757 int dev_change_xdp_fd(struct net_device *dev, struct netlink_ext_ack *extack, 3758 int fd, int expected_fd, u32 flags); 3759 int bpf_xdp_link_attach(const union bpf_attr *attr, struct bpf_prog *prog); 3760 u8 dev_xdp_prog_count(struct net_device *dev); 3761 u32 dev_xdp_prog_id(struct net_device *dev, enum bpf_xdp_mode mode); 3762 3763 int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb); 3764 int dev_forward_skb(struct net_device *dev, struct sk_buff *skb); 3765 int dev_forward_skb_nomtu(struct net_device *dev, struct sk_buff *skb); 3766 bool is_skb_forwardable(const struct net_device *dev, 3767 const struct sk_buff *skb); 3768 3769 static __always_inline bool __is_skb_forwardable(const struct net_device *dev, 3770 const struct sk_buff *skb, 3771 const bool check_mtu) 3772 { 3773 const u32 vlan_hdr_len = 4; /* VLAN_HLEN */ 3774 unsigned int len; 3775 3776 if (!(dev->flags & IFF_UP)) 3777 return false; 3778 3779 if (!check_mtu) 3780 return true; 3781 3782 len = dev->mtu + dev->hard_header_len + vlan_hdr_len; 3783 if (skb->len <= len) 3784 return true; 3785 3786 /* if TSO is enabled, we don't care about the length as the packet 3787 * could be forwarded without being segmented before 3788 */ 3789 if (skb_is_gso(skb)) 3790 return true; 3791 3792 return false; 3793 } 3794 3795 static __always_inline int ____dev_forward_skb(struct net_device *dev, 3796 struct sk_buff *skb, 3797 const bool check_mtu) 3798 { 3799 if (skb_orphan_frags(skb, GFP_ATOMIC) || 3800 unlikely(!__is_skb_forwardable(dev, skb, check_mtu))) { 3801 atomic_long_inc(&dev->rx_dropped); 3802 kfree_skb(skb); 3803 return NET_RX_DROP; 3804 } 3805 3806 skb_scrub_packet(skb, !net_eq(dev_net(dev), dev_net(skb->dev))); 3807 skb->priority = 0; 3808 return 0; 3809 } 3810 3811 bool dev_nit_active(struct net_device *dev); 3812 void dev_queue_xmit_nit(struct sk_buff *skb, struct net_device *dev); 3813 3814 extern int netdev_budget; 3815 extern unsigned int netdev_budget_usecs; 3816 3817 /* Called by rtnetlink.c:rtnl_unlock() */ 3818 void netdev_run_todo(void); 3819 3820 /** 3821 * dev_put - release reference to device 3822 * @dev: network device 3823 * 3824 * Release reference to device to allow it to be freed. 3825 * Try using dev_put_track() instead. 3826 */ 3827 static inline void dev_put(struct net_device *dev) 3828 { 3829 if (dev) { 3830 #ifdef CONFIG_PCPU_DEV_REFCNT 3831 this_cpu_dec(*dev->pcpu_refcnt); 3832 #else 3833 refcount_dec(&dev->dev_refcnt); 3834 #endif 3835 } 3836 } 3837 3838 /** 3839 * dev_hold - get reference to device 3840 * @dev: network device 3841 * 3842 * Hold reference to device to keep it from being freed. 3843 * Try using dev_hold_track() instead. 3844 */ 3845 static inline void dev_hold(struct net_device *dev) 3846 { 3847 if (dev) { 3848 #ifdef CONFIG_PCPU_DEV_REFCNT 3849 this_cpu_inc(*dev->pcpu_refcnt); 3850 #else 3851 refcount_inc(&dev->dev_refcnt); 3852 #endif 3853 } 3854 } 3855 3856 static inline void netdev_tracker_alloc(struct net_device *dev, 3857 netdevice_tracker *tracker, gfp_t gfp) 3858 { 3859 #ifdef CONFIG_NET_DEV_REFCNT_TRACKER 3860 ref_tracker_alloc(&dev->refcnt_tracker, tracker, gfp); 3861 #endif 3862 } 3863 3864 static inline void netdev_tracker_free(struct net_device *dev, 3865 netdevice_tracker *tracker) 3866 { 3867 #ifdef CONFIG_NET_DEV_REFCNT_TRACKER 3868 ref_tracker_free(&dev->refcnt_tracker, tracker); 3869 #endif 3870 } 3871 3872 static inline void dev_hold_track(struct net_device *dev, 3873 netdevice_tracker *tracker, gfp_t gfp) 3874 { 3875 if (dev) { 3876 dev_hold(dev); 3877 netdev_tracker_alloc(dev, tracker, gfp); 3878 } 3879 } 3880 3881 static inline void dev_put_track(struct net_device *dev, 3882 netdevice_tracker *tracker) 3883 { 3884 if (dev) { 3885 netdev_tracker_free(dev, tracker); 3886 dev_put(dev); 3887 } 3888 } 3889 3890 static inline void dev_replace_track(struct net_device *odev, 3891 struct net_device *ndev, 3892 netdevice_tracker *tracker, 3893 gfp_t gfp) 3894 { 3895 #ifdef CONFIG_NET_DEV_REFCNT_TRACKER 3896 if (odev) 3897 ref_tracker_free(&odev->refcnt_tracker, tracker); 3898 #endif 3899 dev_hold(ndev); 3900 dev_put(odev); 3901 #ifdef CONFIG_NET_DEV_REFCNT_TRACKER 3902 if (ndev) 3903 ref_tracker_alloc(&ndev->refcnt_tracker, tracker, gfp); 3904 #endif 3905 } 3906 3907 /* Carrier loss detection, dial on demand. The functions netif_carrier_on 3908 * and _off may be called from IRQ context, but it is caller 3909 * who is responsible for serialization of these calls. 3910 * 3911 * The name carrier is inappropriate, these functions should really be 3912 * called netif_lowerlayer_*() because they represent the state of any 3913 * kind of lower layer not just hardware media. 3914 */ 3915 3916 void linkwatch_init_dev(struct net_device *dev); 3917 void linkwatch_fire_event(struct net_device *dev); 3918 void linkwatch_forget_dev(struct net_device *dev); 3919 3920 /** 3921 * netif_carrier_ok - test if carrier present 3922 * @dev: network device 3923 * 3924 * Check if carrier is present on device 3925 */ 3926 static inline bool netif_carrier_ok(const struct net_device *dev) 3927 { 3928 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state); 3929 } 3930 3931 unsigned long dev_trans_start(struct net_device *dev); 3932 3933 void __netdev_watchdog_up(struct net_device *dev); 3934 3935 void netif_carrier_on(struct net_device *dev); 3936 void netif_carrier_off(struct net_device *dev); 3937 void netif_carrier_event(struct net_device *dev); 3938 3939 /** 3940 * netif_dormant_on - mark device as dormant. 3941 * @dev: network device 3942 * 3943 * Mark device as dormant (as per RFC2863). 3944 * 3945 * The dormant state indicates that the relevant interface is not 3946 * actually in a condition to pass packets (i.e., it is not 'up') but is 3947 * in a "pending" state, waiting for some external event. For "on- 3948 * demand" interfaces, this new state identifies the situation where the 3949 * interface is waiting for events to place it in the up state. 3950 */ 3951 static inline void netif_dormant_on(struct net_device *dev) 3952 { 3953 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state)) 3954 linkwatch_fire_event(dev); 3955 } 3956 3957 /** 3958 * netif_dormant_off - set device as not dormant. 3959 * @dev: network device 3960 * 3961 * Device is not in dormant state. 3962 */ 3963 static inline void netif_dormant_off(struct net_device *dev) 3964 { 3965 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state)) 3966 linkwatch_fire_event(dev); 3967 } 3968 3969 /** 3970 * netif_dormant - test if device is dormant 3971 * @dev: network device 3972 * 3973 * Check if device is dormant. 3974 */ 3975 static inline bool netif_dormant(const struct net_device *dev) 3976 { 3977 return test_bit(__LINK_STATE_DORMANT, &dev->state); 3978 } 3979 3980 3981 /** 3982 * netif_testing_on - mark device as under test. 3983 * @dev: network device 3984 * 3985 * Mark device as under test (as per RFC2863). 3986 * 3987 * The testing state indicates that some test(s) must be performed on 3988 * the interface. After completion, of the test, the interface state 3989 * will change to up, dormant, or down, as appropriate. 3990 */ 3991 static inline void netif_testing_on(struct net_device *dev) 3992 { 3993 if (!test_and_set_bit(__LINK_STATE_TESTING, &dev->state)) 3994 linkwatch_fire_event(dev); 3995 } 3996 3997 /** 3998 * netif_testing_off - set device as not under test. 3999 * @dev: network device 4000 * 4001 * Device is not in testing state. 4002 */ 4003 static inline void netif_testing_off(struct net_device *dev) 4004 { 4005 if (test_and_clear_bit(__LINK_STATE_TESTING, &dev->state)) 4006 linkwatch_fire_event(dev); 4007 } 4008 4009 /** 4010 * netif_testing - test if device is under test 4011 * @dev: network device 4012 * 4013 * Check if device is under test 4014 */ 4015 static inline bool netif_testing(const struct net_device *dev) 4016 { 4017 return test_bit(__LINK_STATE_TESTING, &dev->state); 4018 } 4019 4020 4021 /** 4022 * netif_oper_up - test if device is operational 4023 * @dev: network device 4024 * 4025 * Check if carrier is operational 4026 */ 4027 static inline bool netif_oper_up(const struct net_device *dev) 4028 { 4029 return (dev->operstate == IF_OPER_UP || 4030 dev->operstate == IF_OPER_UNKNOWN /* backward compat */); 4031 } 4032 4033 /** 4034 * netif_device_present - is device available or removed 4035 * @dev: network device 4036 * 4037 * Check if device has not been removed from system. 4038 */ 4039 static inline bool netif_device_present(const struct net_device *dev) 4040 { 4041 return test_bit(__LINK_STATE_PRESENT, &dev->state); 4042 } 4043 4044 void netif_device_detach(struct net_device *dev); 4045 4046 void netif_device_attach(struct net_device *dev); 4047 4048 /* 4049 * Network interface message level settings 4050 */ 4051 4052 enum { 4053 NETIF_MSG_DRV_BIT, 4054 NETIF_MSG_PROBE_BIT, 4055 NETIF_MSG_LINK_BIT, 4056 NETIF_MSG_TIMER_BIT, 4057 NETIF_MSG_IFDOWN_BIT, 4058 NETIF_MSG_IFUP_BIT, 4059 NETIF_MSG_RX_ERR_BIT, 4060 NETIF_MSG_TX_ERR_BIT, 4061 NETIF_MSG_TX_QUEUED_BIT, 4062 NETIF_MSG_INTR_BIT, 4063 NETIF_MSG_TX_DONE_BIT, 4064 NETIF_MSG_RX_STATUS_BIT, 4065 NETIF_MSG_PKTDATA_BIT, 4066 NETIF_MSG_HW_BIT, 4067 NETIF_MSG_WOL_BIT, 4068 4069 /* When you add a new bit above, update netif_msg_class_names array 4070 * in net/ethtool/common.c 4071 */ 4072 NETIF_MSG_CLASS_COUNT, 4073 }; 4074 /* Both ethtool_ops interface and internal driver implementation use u32 */ 4075 static_assert(NETIF_MSG_CLASS_COUNT <= 32); 4076 4077 #define __NETIF_MSG_BIT(bit) ((u32)1 << (bit)) 4078 #define __NETIF_MSG(name) __NETIF_MSG_BIT(NETIF_MSG_ ## name ## _BIT) 4079 4080 #define NETIF_MSG_DRV __NETIF_MSG(DRV) 4081 #define NETIF_MSG_PROBE __NETIF_MSG(PROBE) 4082 #define NETIF_MSG_LINK __NETIF_MSG(LINK) 4083 #define NETIF_MSG_TIMER __NETIF_MSG(TIMER) 4084 #define NETIF_MSG_IFDOWN __NETIF_MSG(IFDOWN) 4085 #define NETIF_MSG_IFUP __NETIF_MSG(IFUP) 4086 #define NETIF_MSG_RX_ERR __NETIF_MSG(RX_ERR) 4087 #define NETIF_MSG_TX_ERR __NETIF_MSG(TX_ERR) 4088 #define NETIF_MSG_TX_QUEUED __NETIF_MSG(TX_QUEUED) 4089 #define NETIF_MSG_INTR __NETIF_MSG(INTR) 4090 #define NETIF_MSG_TX_DONE __NETIF_MSG(TX_DONE) 4091 #define NETIF_MSG_RX_STATUS __NETIF_MSG(RX_STATUS) 4092 #define NETIF_MSG_PKTDATA __NETIF_MSG(PKTDATA) 4093 #define NETIF_MSG_HW __NETIF_MSG(HW) 4094 #define NETIF_MSG_WOL __NETIF_MSG(WOL) 4095 4096 #define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV) 4097 #define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE) 4098 #define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK) 4099 #define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER) 4100 #define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN) 4101 #define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP) 4102 #define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR) 4103 #define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR) 4104 #define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED) 4105 #define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR) 4106 #define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE) 4107 #define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS) 4108 #define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA) 4109 #define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW) 4110 #define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL) 4111 4112 static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits) 4113 { 4114 /* use default */ 4115 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8)) 4116 return default_msg_enable_bits; 4117 if (debug_value == 0) /* no output */ 4118 return 0; 4119 /* set low N bits */ 4120 return (1U << debug_value) - 1; 4121 } 4122 4123 static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu) 4124 { 4125 spin_lock(&txq->_xmit_lock); 4126 /* Pairs with READ_ONCE() in __dev_queue_xmit() */ 4127 WRITE_ONCE(txq->xmit_lock_owner, cpu); 4128 } 4129 4130 static inline bool __netif_tx_acquire(struct netdev_queue *txq) 4131 { 4132 __acquire(&txq->_xmit_lock); 4133 return true; 4134 } 4135 4136 static inline void __netif_tx_release(struct netdev_queue *txq) 4137 { 4138 __release(&txq->_xmit_lock); 4139 } 4140 4141 static inline void __netif_tx_lock_bh(struct netdev_queue *txq) 4142 { 4143 spin_lock_bh(&txq->_xmit_lock); 4144 /* Pairs with READ_ONCE() in __dev_queue_xmit() */ 4145 WRITE_ONCE(txq->xmit_lock_owner, smp_processor_id()); 4146 } 4147 4148 static inline bool __netif_tx_trylock(struct netdev_queue *txq) 4149 { 4150 bool ok = spin_trylock(&txq->_xmit_lock); 4151 4152 if (likely(ok)) { 4153 /* Pairs with READ_ONCE() in __dev_queue_xmit() */ 4154 WRITE_ONCE(txq->xmit_lock_owner, smp_processor_id()); 4155 } 4156 return ok; 4157 } 4158 4159 static inline void __netif_tx_unlock(struct netdev_queue *txq) 4160 { 4161 /* Pairs with READ_ONCE() in __dev_queue_xmit() */ 4162 WRITE_ONCE(txq->xmit_lock_owner, -1); 4163 spin_unlock(&txq->_xmit_lock); 4164 } 4165 4166 static inline void __netif_tx_unlock_bh(struct netdev_queue *txq) 4167 { 4168 /* Pairs with READ_ONCE() in __dev_queue_xmit() */ 4169 WRITE_ONCE(txq->xmit_lock_owner, -1); 4170 spin_unlock_bh(&txq->_xmit_lock); 4171 } 4172 4173 /* 4174 * txq->trans_start can be read locklessly from dev_watchdog() 4175 */ 4176 static inline void txq_trans_update(struct netdev_queue *txq) 4177 { 4178 if (txq->xmit_lock_owner != -1) 4179 WRITE_ONCE(txq->trans_start, jiffies); 4180 } 4181 4182 static inline void txq_trans_cond_update(struct netdev_queue *txq) 4183 { 4184 unsigned long now = jiffies; 4185 4186 if (READ_ONCE(txq->trans_start) != now) 4187 WRITE_ONCE(txq->trans_start, now); 4188 } 4189 4190 /* legacy drivers only, netdev_start_xmit() sets txq->trans_start */ 4191 static inline void netif_trans_update(struct net_device *dev) 4192 { 4193 struct netdev_queue *txq = netdev_get_tx_queue(dev, 0); 4194 4195 txq_trans_cond_update(txq); 4196 } 4197 4198 /** 4199 * netif_tx_lock - grab network device transmit lock 4200 * @dev: network device 4201 * 4202 * Get network device transmit lock 4203 */ 4204 void netif_tx_lock(struct net_device *dev); 4205 4206 static inline void netif_tx_lock_bh(struct net_device *dev) 4207 { 4208 local_bh_disable(); 4209 netif_tx_lock(dev); 4210 } 4211 4212 void netif_tx_unlock(struct net_device *dev); 4213 4214 static inline void netif_tx_unlock_bh(struct net_device *dev) 4215 { 4216 netif_tx_unlock(dev); 4217 local_bh_enable(); 4218 } 4219 4220 #define HARD_TX_LOCK(dev, txq, cpu) { \ 4221 if ((dev->features & NETIF_F_LLTX) == 0) { \ 4222 __netif_tx_lock(txq, cpu); \ 4223 } else { \ 4224 __netif_tx_acquire(txq); \ 4225 } \ 4226 } 4227 4228 #define HARD_TX_TRYLOCK(dev, txq) \ 4229 (((dev->features & NETIF_F_LLTX) == 0) ? \ 4230 __netif_tx_trylock(txq) : \ 4231 __netif_tx_acquire(txq)) 4232 4233 #define HARD_TX_UNLOCK(dev, txq) { \ 4234 if ((dev->features & NETIF_F_LLTX) == 0) { \ 4235 __netif_tx_unlock(txq); \ 4236 } else { \ 4237 __netif_tx_release(txq); \ 4238 } \ 4239 } 4240 4241 static inline void netif_tx_disable(struct net_device *dev) 4242 { 4243 unsigned int i; 4244 int cpu; 4245 4246 local_bh_disable(); 4247 cpu = smp_processor_id(); 4248 spin_lock(&dev->tx_global_lock); 4249 for (i = 0; i < dev->num_tx_queues; i++) { 4250 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 4251 4252 __netif_tx_lock(txq, cpu); 4253 netif_tx_stop_queue(txq); 4254 __netif_tx_unlock(txq); 4255 } 4256 spin_unlock(&dev->tx_global_lock); 4257 local_bh_enable(); 4258 } 4259 4260 static inline void netif_addr_lock(struct net_device *dev) 4261 { 4262 unsigned char nest_level = 0; 4263 4264 #ifdef CONFIG_LOCKDEP 4265 nest_level = dev->nested_level; 4266 #endif 4267 spin_lock_nested(&dev->addr_list_lock, nest_level); 4268 } 4269 4270 static inline void netif_addr_lock_bh(struct net_device *dev) 4271 { 4272 unsigned char nest_level = 0; 4273 4274 #ifdef CONFIG_LOCKDEP 4275 nest_level = dev->nested_level; 4276 #endif 4277 local_bh_disable(); 4278 spin_lock_nested(&dev->addr_list_lock, nest_level); 4279 } 4280 4281 static inline void netif_addr_unlock(struct net_device *dev) 4282 { 4283 spin_unlock(&dev->addr_list_lock); 4284 } 4285 4286 static inline void netif_addr_unlock_bh(struct net_device *dev) 4287 { 4288 spin_unlock_bh(&dev->addr_list_lock); 4289 } 4290 4291 /* 4292 * dev_addrs walker. Should be used only for read access. Call with 4293 * rcu_read_lock held. 4294 */ 4295 #define for_each_dev_addr(dev, ha) \ 4296 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list) 4297 4298 /* These functions live elsewhere (drivers/net/net_init.c, but related) */ 4299 4300 void ether_setup(struct net_device *dev); 4301 4302 /* Support for loadable net-drivers */ 4303 struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name, 4304 unsigned char name_assign_type, 4305 void (*setup)(struct net_device *), 4306 unsigned int txqs, unsigned int rxqs); 4307 #define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \ 4308 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1) 4309 4310 #define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \ 4311 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \ 4312 count) 4313 4314 int register_netdev(struct net_device *dev); 4315 void unregister_netdev(struct net_device *dev); 4316 4317 int devm_register_netdev(struct device *dev, struct net_device *ndev); 4318 4319 /* General hardware address lists handling functions */ 4320 int __hw_addr_sync(struct netdev_hw_addr_list *to_list, 4321 struct netdev_hw_addr_list *from_list, int addr_len); 4322 void __hw_addr_unsync(struct netdev_hw_addr_list *to_list, 4323 struct netdev_hw_addr_list *from_list, int addr_len); 4324 int __hw_addr_sync_dev(struct netdev_hw_addr_list *list, 4325 struct net_device *dev, 4326 int (*sync)(struct net_device *, const unsigned char *), 4327 int (*unsync)(struct net_device *, 4328 const unsigned char *)); 4329 int __hw_addr_ref_sync_dev(struct netdev_hw_addr_list *list, 4330 struct net_device *dev, 4331 int (*sync)(struct net_device *, 4332 const unsigned char *, int), 4333 int (*unsync)(struct net_device *, 4334 const unsigned char *, int)); 4335 void __hw_addr_ref_unsync_dev(struct netdev_hw_addr_list *list, 4336 struct net_device *dev, 4337 int (*unsync)(struct net_device *, 4338 const unsigned char *, int)); 4339 void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list, 4340 struct net_device *dev, 4341 int (*unsync)(struct net_device *, 4342 const unsigned char *)); 4343 void __hw_addr_init(struct netdev_hw_addr_list *list); 4344 4345 /* Functions used for device addresses handling */ 4346 void dev_addr_mod(struct net_device *dev, unsigned int offset, 4347 const void *addr, size_t len); 4348 4349 static inline void 4350 __dev_addr_set(struct net_device *dev, const void *addr, size_t len) 4351 { 4352 dev_addr_mod(dev, 0, addr, len); 4353 } 4354 4355 static inline void dev_addr_set(struct net_device *dev, const u8 *addr) 4356 { 4357 __dev_addr_set(dev, addr, dev->addr_len); 4358 } 4359 4360 int dev_addr_add(struct net_device *dev, const unsigned char *addr, 4361 unsigned char addr_type); 4362 int dev_addr_del(struct net_device *dev, const unsigned char *addr, 4363 unsigned char addr_type); 4364 void dev_addr_flush(struct net_device *dev); 4365 int dev_addr_init(struct net_device *dev); 4366 void dev_addr_check(struct net_device *dev); 4367 4368 /* Functions used for unicast addresses handling */ 4369 int dev_uc_add(struct net_device *dev, const unsigned char *addr); 4370 int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr); 4371 int dev_uc_del(struct net_device *dev, const unsigned char *addr); 4372 int dev_uc_sync(struct net_device *to, struct net_device *from); 4373 int dev_uc_sync_multiple(struct net_device *to, struct net_device *from); 4374 void dev_uc_unsync(struct net_device *to, struct net_device *from); 4375 void dev_uc_flush(struct net_device *dev); 4376 void dev_uc_init(struct net_device *dev); 4377 4378 /** 4379 * __dev_uc_sync - Synchonize device's unicast list 4380 * @dev: device to sync 4381 * @sync: function to call if address should be added 4382 * @unsync: function to call if address should be removed 4383 * 4384 * Add newly added addresses to the interface, and release 4385 * addresses that have been deleted. 4386 */ 4387 static inline int __dev_uc_sync(struct net_device *dev, 4388 int (*sync)(struct net_device *, 4389 const unsigned char *), 4390 int (*unsync)(struct net_device *, 4391 const unsigned char *)) 4392 { 4393 return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync); 4394 } 4395 4396 /** 4397 * __dev_uc_unsync - Remove synchronized addresses from device 4398 * @dev: device to sync 4399 * @unsync: function to call if address should be removed 4400 * 4401 * Remove all addresses that were added to the device by dev_uc_sync(). 4402 */ 4403 static inline void __dev_uc_unsync(struct net_device *dev, 4404 int (*unsync)(struct net_device *, 4405 const unsigned char *)) 4406 { 4407 __hw_addr_unsync_dev(&dev->uc, dev, unsync); 4408 } 4409 4410 /* Functions used for multicast addresses handling */ 4411 int dev_mc_add(struct net_device *dev, const unsigned char *addr); 4412 int dev_mc_add_global(struct net_device *dev, const unsigned char *addr); 4413 int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr); 4414 int dev_mc_del(struct net_device *dev, const unsigned char *addr); 4415 int dev_mc_del_global(struct net_device *dev, const unsigned char *addr); 4416 int dev_mc_sync(struct net_device *to, struct net_device *from); 4417 int dev_mc_sync_multiple(struct net_device *to, struct net_device *from); 4418 void dev_mc_unsync(struct net_device *to, struct net_device *from); 4419 void dev_mc_flush(struct net_device *dev); 4420 void dev_mc_init(struct net_device *dev); 4421 4422 /** 4423 * __dev_mc_sync - Synchonize device's multicast list 4424 * @dev: device to sync 4425 * @sync: function to call if address should be added 4426 * @unsync: function to call if address should be removed 4427 * 4428 * Add newly added addresses to the interface, and release 4429 * addresses that have been deleted. 4430 */ 4431 static inline int __dev_mc_sync(struct net_device *dev, 4432 int (*sync)(struct net_device *, 4433 const unsigned char *), 4434 int (*unsync)(struct net_device *, 4435 const unsigned char *)) 4436 { 4437 return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync); 4438 } 4439 4440 /** 4441 * __dev_mc_unsync - Remove synchronized addresses from device 4442 * @dev: device to sync 4443 * @unsync: function to call if address should be removed 4444 * 4445 * Remove all addresses that were added to the device by dev_mc_sync(). 4446 */ 4447 static inline void __dev_mc_unsync(struct net_device *dev, 4448 int (*unsync)(struct net_device *, 4449 const unsigned char *)) 4450 { 4451 __hw_addr_unsync_dev(&dev->mc, dev, unsync); 4452 } 4453 4454 /* Functions used for secondary unicast and multicast support */ 4455 void dev_set_rx_mode(struct net_device *dev); 4456 void __dev_set_rx_mode(struct net_device *dev); 4457 int dev_set_promiscuity(struct net_device *dev, int inc); 4458 int dev_set_allmulti(struct net_device *dev, int inc); 4459 void netdev_state_change(struct net_device *dev); 4460 void __netdev_notify_peers(struct net_device *dev); 4461 void netdev_notify_peers(struct net_device *dev); 4462 void netdev_features_change(struct net_device *dev); 4463 /* Load a device via the kmod */ 4464 void dev_load(struct net *net, const char *name); 4465 struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev, 4466 struct rtnl_link_stats64 *storage); 4467 void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64, 4468 const struct net_device_stats *netdev_stats); 4469 void dev_fetch_sw_netstats(struct rtnl_link_stats64 *s, 4470 const struct pcpu_sw_netstats __percpu *netstats); 4471 void dev_get_tstats64(struct net_device *dev, struct rtnl_link_stats64 *s); 4472 4473 extern int netdev_max_backlog; 4474 extern int netdev_tstamp_prequeue; 4475 extern int netdev_unregister_timeout_secs; 4476 extern int weight_p; 4477 extern int dev_weight_rx_bias; 4478 extern int dev_weight_tx_bias; 4479 extern int dev_rx_weight; 4480 extern int dev_tx_weight; 4481 extern int gro_normal_batch; 4482 4483 enum { 4484 NESTED_SYNC_IMM_BIT, 4485 NESTED_SYNC_TODO_BIT, 4486 }; 4487 4488 #define __NESTED_SYNC_BIT(bit) ((u32)1 << (bit)) 4489 #define __NESTED_SYNC(name) __NESTED_SYNC_BIT(NESTED_SYNC_ ## name ## _BIT) 4490 4491 #define NESTED_SYNC_IMM __NESTED_SYNC(IMM) 4492 #define NESTED_SYNC_TODO __NESTED_SYNC(TODO) 4493 4494 struct netdev_nested_priv { 4495 unsigned char flags; 4496 void *data; 4497 }; 4498 4499 bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev); 4500 struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev, 4501 struct list_head **iter); 4502 4503 #ifdef CONFIG_LOCKDEP 4504 static LIST_HEAD(net_unlink_list); 4505 4506 static inline void net_unlink_todo(struct net_device *dev) 4507 { 4508 if (list_empty(&dev->unlink_list)) 4509 list_add_tail(&dev->unlink_list, &net_unlink_list); 4510 } 4511 #endif 4512 4513 /* iterate through upper list, must be called under RCU read lock */ 4514 #define netdev_for_each_upper_dev_rcu(dev, updev, iter) \ 4515 for (iter = &(dev)->adj_list.upper, \ 4516 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \ 4517 updev; \ 4518 updev = netdev_upper_get_next_dev_rcu(dev, &(iter))) 4519 4520 int netdev_walk_all_upper_dev_rcu(struct net_device *dev, 4521 int (*fn)(struct net_device *upper_dev, 4522 struct netdev_nested_priv *priv), 4523 struct netdev_nested_priv *priv); 4524 4525 bool netdev_has_upper_dev_all_rcu(struct net_device *dev, 4526 struct net_device *upper_dev); 4527 4528 bool netdev_has_any_upper_dev(struct net_device *dev); 4529 4530 void *netdev_lower_get_next_private(struct net_device *dev, 4531 struct list_head **iter); 4532 void *netdev_lower_get_next_private_rcu(struct net_device *dev, 4533 struct list_head **iter); 4534 4535 #define netdev_for_each_lower_private(dev, priv, iter) \ 4536 for (iter = (dev)->adj_list.lower.next, \ 4537 priv = netdev_lower_get_next_private(dev, &(iter)); \ 4538 priv; \ 4539 priv = netdev_lower_get_next_private(dev, &(iter))) 4540 4541 #define netdev_for_each_lower_private_rcu(dev, priv, iter) \ 4542 for (iter = &(dev)->adj_list.lower, \ 4543 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \ 4544 priv; \ 4545 priv = netdev_lower_get_next_private_rcu(dev, &(iter))) 4546 4547 void *netdev_lower_get_next(struct net_device *dev, 4548 struct list_head **iter); 4549 4550 #define netdev_for_each_lower_dev(dev, ldev, iter) \ 4551 for (iter = (dev)->adj_list.lower.next, \ 4552 ldev = netdev_lower_get_next(dev, &(iter)); \ 4553 ldev; \ 4554 ldev = netdev_lower_get_next(dev, &(iter))) 4555 4556 struct net_device *netdev_next_lower_dev_rcu(struct net_device *dev, 4557 struct list_head **iter); 4558 int netdev_walk_all_lower_dev(struct net_device *dev, 4559 int (*fn)(struct net_device *lower_dev, 4560 struct netdev_nested_priv *priv), 4561 struct netdev_nested_priv *priv); 4562 int netdev_walk_all_lower_dev_rcu(struct net_device *dev, 4563 int (*fn)(struct net_device *lower_dev, 4564 struct netdev_nested_priv *priv), 4565 struct netdev_nested_priv *priv); 4566 4567 void *netdev_adjacent_get_private(struct list_head *adj_list); 4568 void *netdev_lower_get_first_private_rcu(struct net_device *dev); 4569 struct net_device *netdev_master_upper_dev_get(struct net_device *dev); 4570 struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev); 4571 int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev, 4572 struct netlink_ext_ack *extack); 4573 int netdev_master_upper_dev_link(struct net_device *dev, 4574 struct net_device *upper_dev, 4575 void *upper_priv, void *upper_info, 4576 struct netlink_ext_ack *extack); 4577 void netdev_upper_dev_unlink(struct net_device *dev, 4578 struct net_device *upper_dev); 4579 int netdev_adjacent_change_prepare(struct net_device *old_dev, 4580 struct net_device *new_dev, 4581 struct net_device *dev, 4582 struct netlink_ext_ack *extack); 4583 void netdev_adjacent_change_commit(struct net_device *old_dev, 4584 struct net_device *new_dev, 4585 struct net_device *dev); 4586 void netdev_adjacent_change_abort(struct net_device *old_dev, 4587 struct net_device *new_dev, 4588 struct net_device *dev); 4589 void netdev_adjacent_rename_links(struct net_device *dev, char *oldname); 4590 void *netdev_lower_dev_get_private(struct net_device *dev, 4591 struct net_device *lower_dev); 4592 void netdev_lower_state_changed(struct net_device *lower_dev, 4593 void *lower_state_info); 4594 4595 /* RSS keys are 40 or 52 bytes long */ 4596 #define NETDEV_RSS_KEY_LEN 52 4597 extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN] __read_mostly; 4598 void netdev_rss_key_fill(void *buffer, size_t len); 4599 4600 int skb_checksum_help(struct sk_buff *skb); 4601 int skb_crc32c_csum_help(struct sk_buff *skb); 4602 int skb_csum_hwoffload_help(struct sk_buff *skb, 4603 const netdev_features_t features); 4604 4605 struct sk_buff *__skb_gso_segment(struct sk_buff *skb, 4606 netdev_features_t features, bool tx_path); 4607 struct sk_buff *skb_mac_gso_segment(struct sk_buff *skb, 4608 netdev_features_t features); 4609 4610 struct netdev_bonding_info { 4611 ifslave slave; 4612 ifbond master; 4613 }; 4614 4615 struct netdev_notifier_bonding_info { 4616 struct netdev_notifier_info info; /* must be first */ 4617 struct netdev_bonding_info bonding_info; 4618 }; 4619 4620 void netdev_bonding_info_change(struct net_device *dev, 4621 struct netdev_bonding_info *bonding_info); 4622 4623 #if IS_ENABLED(CONFIG_ETHTOOL_NETLINK) 4624 void ethtool_notify(struct net_device *dev, unsigned int cmd, const void *data); 4625 #else 4626 static inline void ethtool_notify(struct net_device *dev, unsigned int cmd, 4627 const void *data) 4628 { 4629 } 4630 #endif 4631 4632 static inline 4633 struct sk_buff *skb_gso_segment(struct sk_buff *skb, netdev_features_t features) 4634 { 4635 return __skb_gso_segment(skb, features, true); 4636 } 4637 __be16 skb_network_protocol(struct sk_buff *skb, int *depth); 4638 4639 static inline bool can_checksum_protocol(netdev_features_t features, 4640 __be16 protocol) 4641 { 4642 if (protocol == htons(ETH_P_FCOE)) 4643 return !!(features & NETIF_F_FCOE_CRC); 4644 4645 /* Assume this is an IP checksum (not SCTP CRC) */ 4646 4647 if (features & NETIF_F_HW_CSUM) { 4648 /* Can checksum everything */ 4649 return true; 4650 } 4651 4652 switch (protocol) { 4653 case htons(ETH_P_IP): 4654 return !!(features & NETIF_F_IP_CSUM); 4655 case htons(ETH_P_IPV6): 4656 return !!(features & NETIF_F_IPV6_CSUM); 4657 default: 4658 return false; 4659 } 4660 } 4661 4662 #ifdef CONFIG_BUG 4663 void netdev_rx_csum_fault(struct net_device *dev, struct sk_buff *skb); 4664 #else 4665 static inline void netdev_rx_csum_fault(struct net_device *dev, 4666 struct sk_buff *skb) 4667 { 4668 } 4669 #endif 4670 /* rx skb timestamps */ 4671 void net_enable_timestamp(void); 4672 void net_disable_timestamp(void); 4673 4674 #ifdef CONFIG_PROC_FS 4675 int __init dev_proc_init(void); 4676 #else 4677 #define dev_proc_init() 0 4678 #endif 4679 4680 static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops, 4681 struct sk_buff *skb, struct net_device *dev, 4682 bool more) 4683 { 4684 __this_cpu_write(softnet_data.xmit.more, more); 4685 return ops->ndo_start_xmit(skb, dev); 4686 } 4687 4688 static inline bool netdev_xmit_more(void) 4689 { 4690 return __this_cpu_read(softnet_data.xmit.more); 4691 } 4692 4693 static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev, 4694 struct netdev_queue *txq, bool more) 4695 { 4696 const struct net_device_ops *ops = dev->netdev_ops; 4697 netdev_tx_t rc; 4698 4699 rc = __netdev_start_xmit(ops, skb, dev, more); 4700 if (rc == NETDEV_TX_OK) 4701 txq_trans_update(txq); 4702 4703 return rc; 4704 } 4705 4706 int netdev_class_create_file_ns(const struct class_attribute *class_attr, 4707 const void *ns); 4708 void netdev_class_remove_file_ns(const struct class_attribute *class_attr, 4709 const void *ns); 4710 4711 extern const struct kobj_ns_type_operations net_ns_type_operations; 4712 4713 const char *netdev_drivername(const struct net_device *dev); 4714 4715 void linkwatch_run_queue(void); 4716 4717 static inline netdev_features_t netdev_intersect_features(netdev_features_t f1, 4718 netdev_features_t f2) 4719 { 4720 if ((f1 ^ f2) & NETIF_F_HW_CSUM) { 4721 if (f1 & NETIF_F_HW_CSUM) 4722 f1 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM); 4723 else 4724 f2 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM); 4725 } 4726 4727 return f1 & f2; 4728 } 4729 4730 static inline netdev_features_t netdev_get_wanted_features( 4731 struct net_device *dev) 4732 { 4733 return (dev->features & ~dev->hw_features) | dev->wanted_features; 4734 } 4735 netdev_features_t netdev_increment_features(netdev_features_t all, 4736 netdev_features_t one, netdev_features_t mask); 4737 4738 /* Allow TSO being used on stacked device : 4739 * Performing the GSO segmentation before last device 4740 * is a performance improvement. 4741 */ 4742 static inline netdev_features_t netdev_add_tso_features(netdev_features_t features, 4743 netdev_features_t mask) 4744 { 4745 return netdev_increment_features(features, NETIF_F_ALL_TSO, mask); 4746 } 4747 4748 int __netdev_update_features(struct net_device *dev); 4749 void netdev_update_features(struct net_device *dev); 4750 void netdev_change_features(struct net_device *dev); 4751 4752 void netif_stacked_transfer_operstate(const struct net_device *rootdev, 4753 struct net_device *dev); 4754 4755 netdev_features_t passthru_features_check(struct sk_buff *skb, 4756 struct net_device *dev, 4757 netdev_features_t features); 4758 netdev_features_t netif_skb_features(struct sk_buff *skb); 4759 4760 static inline bool net_gso_ok(netdev_features_t features, int gso_type) 4761 { 4762 netdev_features_t feature = (netdev_features_t)gso_type << NETIF_F_GSO_SHIFT; 4763 4764 /* check flags correspondence */ 4765 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT)); 4766 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT)); 4767 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT)); 4768 BUILD_BUG_ON(SKB_GSO_TCP_FIXEDID != (NETIF_F_TSO_MANGLEID >> NETIF_F_GSO_SHIFT)); 4769 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT)); 4770 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT)); 4771 BUILD_BUG_ON(SKB_GSO_GRE != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT)); 4772 BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT)); 4773 BUILD_BUG_ON(SKB_GSO_IPXIP4 != (NETIF_F_GSO_IPXIP4 >> NETIF_F_GSO_SHIFT)); 4774 BUILD_BUG_ON(SKB_GSO_IPXIP6 != (NETIF_F_GSO_IPXIP6 >> NETIF_F_GSO_SHIFT)); 4775 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT)); 4776 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT)); 4777 BUILD_BUG_ON(SKB_GSO_PARTIAL != (NETIF_F_GSO_PARTIAL >> NETIF_F_GSO_SHIFT)); 4778 BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT)); 4779 BUILD_BUG_ON(SKB_GSO_SCTP != (NETIF_F_GSO_SCTP >> NETIF_F_GSO_SHIFT)); 4780 BUILD_BUG_ON(SKB_GSO_ESP != (NETIF_F_GSO_ESP >> NETIF_F_GSO_SHIFT)); 4781 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_GSO_UDP >> NETIF_F_GSO_SHIFT)); 4782 BUILD_BUG_ON(SKB_GSO_UDP_L4 != (NETIF_F_GSO_UDP_L4 >> NETIF_F_GSO_SHIFT)); 4783 BUILD_BUG_ON(SKB_GSO_FRAGLIST != (NETIF_F_GSO_FRAGLIST >> NETIF_F_GSO_SHIFT)); 4784 4785 return (features & feature) == feature; 4786 } 4787 4788 static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features) 4789 { 4790 return net_gso_ok(features, skb_shinfo(skb)->gso_type) && 4791 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST)); 4792 } 4793 4794 static inline bool netif_needs_gso(struct sk_buff *skb, 4795 netdev_features_t features) 4796 { 4797 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) || 4798 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) && 4799 (skb->ip_summed != CHECKSUM_UNNECESSARY))); 4800 } 4801 4802 static inline void netif_set_gso_max_size(struct net_device *dev, 4803 unsigned int size) 4804 { 4805 /* dev->gso_max_size is read locklessly from sk_setup_caps() */ 4806 WRITE_ONCE(dev->gso_max_size, size); 4807 } 4808 4809 static inline void netif_set_gso_max_segs(struct net_device *dev, 4810 unsigned int segs) 4811 { 4812 /* dev->gso_max_segs is read locklessly from sk_setup_caps() */ 4813 WRITE_ONCE(dev->gso_max_segs, segs); 4814 } 4815 4816 static inline void skb_gso_error_unwind(struct sk_buff *skb, __be16 protocol, 4817 int pulled_hlen, u16 mac_offset, 4818 int mac_len) 4819 { 4820 skb->protocol = protocol; 4821 skb->encapsulation = 1; 4822 skb_push(skb, pulled_hlen); 4823 skb_reset_transport_header(skb); 4824 skb->mac_header = mac_offset; 4825 skb->network_header = skb->mac_header + mac_len; 4826 skb->mac_len = mac_len; 4827 } 4828 4829 static inline bool netif_is_macsec(const struct net_device *dev) 4830 { 4831 return dev->priv_flags & IFF_MACSEC; 4832 } 4833 4834 static inline bool netif_is_macvlan(const struct net_device *dev) 4835 { 4836 return dev->priv_flags & IFF_MACVLAN; 4837 } 4838 4839 static inline bool netif_is_macvlan_port(const struct net_device *dev) 4840 { 4841 return dev->priv_flags & IFF_MACVLAN_PORT; 4842 } 4843 4844 static inline bool netif_is_bond_master(const struct net_device *dev) 4845 { 4846 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING; 4847 } 4848 4849 static inline bool netif_is_bond_slave(const struct net_device *dev) 4850 { 4851 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING; 4852 } 4853 4854 static inline bool netif_supports_nofcs(struct net_device *dev) 4855 { 4856 return dev->priv_flags & IFF_SUPP_NOFCS; 4857 } 4858 4859 static inline bool netif_has_l3_rx_handler(const struct net_device *dev) 4860 { 4861 return dev->priv_flags & IFF_L3MDEV_RX_HANDLER; 4862 } 4863 4864 static inline bool netif_is_l3_master(const struct net_device *dev) 4865 { 4866 return dev->priv_flags & IFF_L3MDEV_MASTER; 4867 } 4868 4869 static inline bool netif_is_l3_slave(const struct net_device *dev) 4870 { 4871 return dev->priv_flags & IFF_L3MDEV_SLAVE; 4872 } 4873 4874 static inline bool netif_is_bridge_master(const struct net_device *dev) 4875 { 4876 return dev->priv_flags & IFF_EBRIDGE; 4877 } 4878 4879 static inline bool netif_is_bridge_port(const struct net_device *dev) 4880 { 4881 return dev->priv_flags & IFF_BRIDGE_PORT; 4882 } 4883 4884 static inline bool netif_is_ovs_master(const struct net_device *dev) 4885 { 4886 return dev->priv_flags & IFF_OPENVSWITCH; 4887 } 4888 4889 static inline bool netif_is_ovs_port(const struct net_device *dev) 4890 { 4891 return dev->priv_flags & IFF_OVS_DATAPATH; 4892 } 4893 4894 static inline bool netif_is_any_bridge_port(const struct net_device *dev) 4895 { 4896 return netif_is_bridge_port(dev) || netif_is_ovs_port(dev); 4897 } 4898 4899 static inline bool netif_is_team_master(const struct net_device *dev) 4900 { 4901 return dev->priv_flags & IFF_TEAM; 4902 } 4903 4904 static inline bool netif_is_team_port(const struct net_device *dev) 4905 { 4906 return dev->priv_flags & IFF_TEAM_PORT; 4907 } 4908 4909 static inline bool netif_is_lag_master(const struct net_device *dev) 4910 { 4911 return netif_is_bond_master(dev) || netif_is_team_master(dev); 4912 } 4913 4914 static inline bool netif_is_lag_port(const struct net_device *dev) 4915 { 4916 return netif_is_bond_slave(dev) || netif_is_team_port(dev); 4917 } 4918 4919 static inline bool netif_is_rxfh_configured(const struct net_device *dev) 4920 { 4921 return dev->priv_flags & IFF_RXFH_CONFIGURED; 4922 } 4923 4924 static inline bool netif_is_failover(const struct net_device *dev) 4925 { 4926 return dev->priv_flags & IFF_FAILOVER; 4927 } 4928 4929 static inline bool netif_is_failover_slave(const struct net_device *dev) 4930 { 4931 return dev->priv_flags & IFF_FAILOVER_SLAVE; 4932 } 4933 4934 /* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */ 4935 static inline void netif_keep_dst(struct net_device *dev) 4936 { 4937 dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM); 4938 } 4939 4940 /* return true if dev can't cope with mtu frames that need vlan tag insertion */ 4941 static inline bool netif_reduces_vlan_mtu(struct net_device *dev) 4942 { 4943 /* TODO: reserve and use an additional IFF bit, if we get more users */ 4944 return netif_is_macsec(dev); 4945 } 4946 4947 extern struct pernet_operations __net_initdata loopback_net_ops; 4948 4949 /* Logging, debugging and troubleshooting/diagnostic helpers. */ 4950 4951 /* netdev_printk helpers, similar to dev_printk */ 4952 4953 static inline const char *netdev_name(const struct net_device *dev) 4954 { 4955 if (!dev->name[0] || strchr(dev->name, '%')) 4956 return "(unnamed net_device)"; 4957 return dev->name; 4958 } 4959 4960 static inline bool netdev_unregistering(const struct net_device *dev) 4961 { 4962 return dev->reg_state == NETREG_UNREGISTERING; 4963 } 4964 4965 static inline const char *netdev_reg_state(const struct net_device *dev) 4966 { 4967 switch (dev->reg_state) { 4968 case NETREG_UNINITIALIZED: return " (uninitialized)"; 4969 case NETREG_REGISTERED: return ""; 4970 case NETREG_UNREGISTERING: return " (unregistering)"; 4971 case NETREG_UNREGISTERED: return " (unregistered)"; 4972 case NETREG_RELEASED: return " (released)"; 4973 case NETREG_DUMMY: return " (dummy)"; 4974 } 4975 4976 WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, dev->reg_state); 4977 return " (unknown)"; 4978 } 4979 4980 __printf(3, 4) __cold 4981 void netdev_printk(const char *level, const struct net_device *dev, 4982 const char *format, ...); 4983 __printf(2, 3) __cold 4984 void netdev_emerg(const struct net_device *dev, const char *format, ...); 4985 __printf(2, 3) __cold 4986 void netdev_alert(const struct net_device *dev, const char *format, ...); 4987 __printf(2, 3) __cold 4988 void netdev_crit(const struct net_device *dev, const char *format, ...); 4989 __printf(2, 3) __cold 4990 void netdev_err(const struct net_device *dev, const char *format, ...); 4991 __printf(2, 3) __cold 4992 void netdev_warn(const struct net_device *dev, const char *format, ...); 4993 __printf(2, 3) __cold 4994 void netdev_notice(const struct net_device *dev, const char *format, ...); 4995 __printf(2, 3) __cold 4996 void netdev_info(const struct net_device *dev, const char *format, ...); 4997 4998 #define netdev_level_once(level, dev, fmt, ...) \ 4999 do { \ 5000 static bool __section(".data.once") __print_once; \ 5001 \ 5002 if (!__print_once) { \ 5003 __print_once = true; \ 5004 netdev_printk(level, dev, fmt, ##__VA_ARGS__); \ 5005 } \ 5006 } while (0) 5007 5008 #define netdev_emerg_once(dev, fmt, ...) \ 5009 netdev_level_once(KERN_EMERG, dev, fmt, ##__VA_ARGS__) 5010 #define netdev_alert_once(dev, fmt, ...) \ 5011 netdev_level_once(KERN_ALERT, dev, fmt, ##__VA_ARGS__) 5012 #define netdev_crit_once(dev, fmt, ...) \ 5013 netdev_level_once(KERN_CRIT, dev, fmt, ##__VA_ARGS__) 5014 #define netdev_err_once(dev, fmt, ...) \ 5015 netdev_level_once(KERN_ERR, dev, fmt, ##__VA_ARGS__) 5016 #define netdev_warn_once(dev, fmt, ...) \ 5017 netdev_level_once(KERN_WARNING, dev, fmt, ##__VA_ARGS__) 5018 #define netdev_notice_once(dev, fmt, ...) \ 5019 netdev_level_once(KERN_NOTICE, dev, fmt, ##__VA_ARGS__) 5020 #define netdev_info_once(dev, fmt, ...) \ 5021 netdev_level_once(KERN_INFO, dev, fmt, ##__VA_ARGS__) 5022 5023 #define MODULE_ALIAS_NETDEV(device) \ 5024 MODULE_ALIAS("netdev-" device) 5025 5026 #if defined(CONFIG_DYNAMIC_DEBUG) || \ 5027 (defined(CONFIG_DYNAMIC_DEBUG_CORE) && defined(DYNAMIC_DEBUG_MODULE)) 5028 #define netdev_dbg(__dev, format, args...) \ 5029 do { \ 5030 dynamic_netdev_dbg(__dev, format, ##args); \ 5031 } while (0) 5032 #elif defined(DEBUG) 5033 #define netdev_dbg(__dev, format, args...) \ 5034 netdev_printk(KERN_DEBUG, __dev, format, ##args) 5035 #else 5036 #define netdev_dbg(__dev, format, args...) \ 5037 ({ \ 5038 if (0) \ 5039 netdev_printk(KERN_DEBUG, __dev, format, ##args); \ 5040 }) 5041 #endif 5042 5043 #if defined(VERBOSE_DEBUG) 5044 #define netdev_vdbg netdev_dbg 5045 #else 5046 5047 #define netdev_vdbg(dev, format, args...) \ 5048 ({ \ 5049 if (0) \ 5050 netdev_printk(KERN_DEBUG, dev, format, ##args); \ 5051 0; \ 5052 }) 5053 #endif 5054 5055 /* 5056 * netdev_WARN() acts like dev_printk(), but with the key difference 5057 * of using a WARN/WARN_ON to get the message out, including the 5058 * file/line information and a backtrace. 5059 */ 5060 #define netdev_WARN(dev, format, args...) \ 5061 WARN(1, "netdevice: %s%s: " format, netdev_name(dev), \ 5062 netdev_reg_state(dev), ##args) 5063 5064 #define netdev_WARN_ONCE(dev, format, args...) \ 5065 WARN_ONCE(1, "netdevice: %s%s: " format, netdev_name(dev), \ 5066 netdev_reg_state(dev), ##args) 5067 5068 /* netif printk helpers, similar to netdev_printk */ 5069 5070 #define netif_printk(priv, type, level, dev, fmt, args...) \ 5071 do { \ 5072 if (netif_msg_##type(priv)) \ 5073 netdev_printk(level, (dev), fmt, ##args); \ 5074 } while (0) 5075 5076 #define netif_level(level, priv, type, dev, fmt, args...) \ 5077 do { \ 5078 if (netif_msg_##type(priv)) \ 5079 netdev_##level(dev, fmt, ##args); \ 5080 } while (0) 5081 5082 #define netif_emerg(priv, type, dev, fmt, args...) \ 5083 netif_level(emerg, priv, type, dev, fmt, ##args) 5084 #define netif_alert(priv, type, dev, fmt, args...) \ 5085 netif_level(alert, priv, type, dev, fmt, ##args) 5086 #define netif_crit(priv, type, dev, fmt, args...) \ 5087 netif_level(crit, priv, type, dev, fmt, ##args) 5088 #define netif_err(priv, type, dev, fmt, args...) \ 5089 netif_level(err, priv, type, dev, fmt, ##args) 5090 #define netif_warn(priv, type, dev, fmt, args...) \ 5091 netif_level(warn, priv, type, dev, fmt, ##args) 5092 #define netif_notice(priv, type, dev, fmt, args...) \ 5093 netif_level(notice, priv, type, dev, fmt, ##args) 5094 #define netif_info(priv, type, dev, fmt, args...) \ 5095 netif_level(info, priv, type, dev, fmt, ##args) 5096 5097 #if defined(CONFIG_DYNAMIC_DEBUG) || \ 5098 (defined(CONFIG_DYNAMIC_DEBUG_CORE) && defined(DYNAMIC_DEBUG_MODULE)) 5099 #define netif_dbg(priv, type, netdev, format, args...) \ 5100 do { \ 5101 if (netif_msg_##type(priv)) \ 5102 dynamic_netdev_dbg(netdev, format, ##args); \ 5103 } while (0) 5104 #elif defined(DEBUG) 5105 #define netif_dbg(priv, type, dev, format, args...) \ 5106 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args) 5107 #else 5108 #define netif_dbg(priv, type, dev, format, args...) \ 5109 ({ \ 5110 if (0) \ 5111 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \ 5112 0; \ 5113 }) 5114 #endif 5115 5116 /* if @cond then downgrade to debug, else print at @level */ 5117 #define netif_cond_dbg(priv, type, netdev, cond, level, fmt, args...) \ 5118 do { \ 5119 if (cond) \ 5120 netif_dbg(priv, type, netdev, fmt, ##args); \ 5121 else \ 5122 netif_ ## level(priv, type, netdev, fmt, ##args); \ 5123 } while (0) 5124 5125 #if defined(VERBOSE_DEBUG) 5126 #define netif_vdbg netif_dbg 5127 #else 5128 #define netif_vdbg(priv, type, dev, format, args...) \ 5129 ({ \ 5130 if (0) \ 5131 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \ 5132 0; \ 5133 }) 5134 #endif 5135 5136 /* 5137 * The list of packet types we will receive (as opposed to discard) 5138 * and the routines to invoke. 5139 * 5140 * Why 16. Because with 16 the only overlap we get on a hash of the 5141 * low nibble of the protocol value is RARP/SNAP/X.25. 5142 * 5143 * 0800 IP 5144 * 0001 802.3 5145 * 0002 AX.25 5146 * 0004 802.2 5147 * 8035 RARP 5148 * 0005 SNAP 5149 * 0805 X.25 5150 * 0806 ARP 5151 * 8137 IPX 5152 * 0009 Localtalk 5153 * 86DD IPv6 5154 */ 5155 #define PTYPE_HASH_SIZE (16) 5156 #define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1) 5157 5158 extern struct list_head ptype_all __read_mostly; 5159 extern struct list_head ptype_base[PTYPE_HASH_SIZE] __read_mostly; 5160 5161 extern struct net_device *blackhole_netdev; 5162 5163 #endif /* _LINUX_NETDEVICE_H */ 5164