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