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 TC_SETUP_ACT, 924 }; 925 926 /* These structures hold the attributes of bpf state that are being passed 927 * to the netdevice through the bpf op. 928 */ 929 enum bpf_netdev_command { 930 /* Set or clear a bpf program used in the earliest stages of packet 931 * rx. The prog will have been loaded as BPF_PROG_TYPE_XDP. The callee 932 * is responsible for calling bpf_prog_put on any old progs that are 933 * stored. In case of error, the callee need not release the new prog 934 * reference, but on success it takes ownership and must bpf_prog_put 935 * when it is no longer used. 936 */ 937 XDP_SETUP_PROG, 938 XDP_SETUP_PROG_HW, 939 /* BPF program for offload callbacks, invoked at program load time. */ 940 BPF_OFFLOAD_MAP_ALLOC, 941 BPF_OFFLOAD_MAP_FREE, 942 XDP_SETUP_XSK_POOL, 943 }; 944 945 struct bpf_prog_offload_ops; 946 struct netlink_ext_ack; 947 struct xdp_umem; 948 struct xdp_dev_bulk_queue; 949 struct bpf_xdp_link; 950 951 enum bpf_xdp_mode { 952 XDP_MODE_SKB = 0, 953 XDP_MODE_DRV = 1, 954 XDP_MODE_HW = 2, 955 __MAX_XDP_MODE 956 }; 957 958 struct bpf_xdp_entity { 959 struct bpf_prog *prog; 960 struct bpf_xdp_link *link; 961 }; 962 963 struct netdev_bpf { 964 enum bpf_netdev_command command; 965 union { 966 /* XDP_SETUP_PROG */ 967 struct { 968 u32 flags; 969 struct bpf_prog *prog; 970 struct netlink_ext_ack *extack; 971 }; 972 /* BPF_OFFLOAD_MAP_ALLOC, BPF_OFFLOAD_MAP_FREE */ 973 struct { 974 struct bpf_offloaded_map *offmap; 975 }; 976 /* XDP_SETUP_XSK_POOL */ 977 struct { 978 struct xsk_buff_pool *pool; 979 u16 queue_id; 980 } xsk; 981 }; 982 }; 983 984 /* Flags for ndo_xsk_wakeup. */ 985 #define XDP_WAKEUP_RX (1 << 0) 986 #define XDP_WAKEUP_TX (1 << 1) 987 988 #ifdef CONFIG_XFRM_OFFLOAD 989 struct xfrmdev_ops { 990 int (*xdo_dev_state_add) (struct xfrm_state *x); 991 void (*xdo_dev_state_delete) (struct xfrm_state *x); 992 void (*xdo_dev_state_free) (struct xfrm_state *x); 993 bool (*xdo_dev_offload_ok) (struct sk_buff *skb, 994 struct xfrm_state *x); 995 void (*xdo_dev_state_advance_esn) (struct xfrm_state *x); 996 }; 997 #endif 998 999 struct dev_ifalias { 1000 struct rcu_head rcuhead; 1001 char ifalias[]; 1002 }; 1003 1004 struct devlink; 1005 struct tlsdev_ops; 1006 1007 struct netdev_name_node { 1008 struct hlist_node hlist; 1009 struct list_head list; 1010 struct net_device *dev; 1011 const char *name; 1012 }; 1013 1014 int netdev_name_node_alt_create(struct net_device *dev, const char *name); 1015 int netdev_name_node_alt_destroy(struct net_device *dev, const char *name); 1016 1017 struct netdev_net_notifier { 1018 struct list_head list; 1019 struct notifier_block *nb; 1020 }; 1021 1022 /* 1023 * This structure defines the management hooks for network devices. 1024 * The following hooks can be defined; unless noted otherwise, they are 1025 * optional and can be filled with a null pointer. 1026 * 1027 * int (*ndo_init)(struct net_device *dev); 1028 * This function is called once when a network device is registered. 1029 * The network device can use this for any late stage initialization 1030 * or semantic validation. It can fail with an error code which will 1031 * be propagated back to register_netdev. 1032 * 1033 * void (*ndo_uninit)(struct net_device *dev); 1034 * This function is called when device is unregistered or when registration 1035 * fails. It is not called if init fails. 1036 * 1037 * int (*ndo_open)(struct net_device *dev); 1038 * This function is called when a network device transitions to the up 1039 * state. 1040 * 1041 * int (*ndo_stop)(struct net_device *dev); 1042 * This function is called when a network device transitions to the down 1043 * state. 1044 * 1045 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb, 1046 * struct net_device *dev); 1047 * Called when a packet needs to be transmitted. 1048 * Returns NETDEV_TX_OK. Can return NETDEV_TX_BUSY, but you should stop 1049 * the queue before that can happen; it's for obsolete devices and weird 1050 * corner cases, but the stack really does a non-trivial amount 1051 * of useless work if you return NETDEV_TX_BUSY. 1052 * Required; cannot be NULL. 1053 * 1054 * netdev_features_t (*ndo_features_check)(struct sk_buff *skb, 1055 * struct net_device *dev 1056 * netdev_features_t features); 1057 * Called by core transmit path to determine if device is capable of 1058 * performing offload operations on a given packet. This is to give 1059 * the device an opportunity to implement any restrictions that cannot 1060 * be otherwise expressed by feature flags. The check is called with 1061 * the set of features that the stack has calculated and it returns 1062 * those the driver believes to be appropriate. 1063 * 1064 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb, 1065 * struct net_device *sb_dev); 1066 * Called to decide which queue to use when device supports multiple 1067 * transmit queues. 1068 * 1069 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags); 1070 * This function is called to allow device receiver to make 1071 * changes to configuration when multicast or promiscuous is enabled. 1072 * 1073 * void (*ndo_set_rx_mode)(struct net_device *dev); 1074 * This function is called device changes address list filtering. 1075 * If driver handles unicast address filtering, it should set 1076 * IFF_UNICAST_FLT in its priv_flags. 1077 * 1078 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr); 1079 * This function is called when the Media Access Control address 1080 * needs to be changed. If this interface is not defined, the 1081 * MAC address can not be changed. 1082 * 1083 * int (*ndo_validate_addr)(struct net_device *dev); 1084 * Test if Media Access Control address is valid for the device. 1085 * 1086 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd); 1087 * Old-style ioctl entry point. This is used internally by the 1088 * appletalk and ieee802154 subsystems but is no longer called by 1089 * the device ioctl handler. 1090 * 1091 * int (*ndo_siocbond)(struct net_device *dev, struct ifreq *ifr, int cmd); 1092 * Used by the bonding driver for its device specific ioctls: 1093 * SIOCBONDENSLAVE, SIOCBONDRELEASE, SIOCBONDSETHWADDR, SIOCBONDCHANGEACTIVE, 1094 * SIOCBONDSLAVEINFOQUERY, and SIOCBONDINFOQUERY 1095 * 1096 * * int (*ndo_eth_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd); 1097 * Called for ethernet specific ioctls: SIOCGMIIPHY, SIOCGMIIREG, 1098 * SIOCSMIIREG, SIOCSHWTSTAMP and SIOCGHWTSTAMP. 1099 * 1100 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map); 1101 * Used to set network devices bus interface parameters. This interface 1102 * is retained for legacy reasons; new devices should use the bus 1103 * interface (PCI) for low level management. 1104 * 1105 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu); 1106 * Called when a user wants to change the Maximum Transfer Unit 1107 * of a device. 1108 * 1109 * void (*ndo_tx_timeout)(struct net_device *dev, unsigned int txqueue); 1110 * Callback used when the transmitter has not made any progress 1111 * for dev->watchdog ticks. 1112 * 1113 * void (*ndo_get_stats64)(struct net_device *dev, 1114 * struct rtnl_link_stats64 *storage); 1115 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev); 1116 * Called when a user wants to get the network device usage 1117 * statistics. Drivers must do one of the following: 1118 * 1. Define @ndo_get_stats64 to fill in a zero-initialised 1119 * rtnl_link_stats64 structure passed by the caller. 1120 * 2. Define @ndo_get_stats to update a net_device_stats structure 1121 * (which should normally be dev->stats) and return a pointer to 1122 * it. The structure may be changed asynchronously only if each 1123 * field is written atomically. 1124 * 3. Update dev->stats asynchronously and atomically, and define 1125 * neither operation. 1126 * 1127 * bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id) 1128 * Return true if this device supports offload stats of this attr_id. 1129 * 1130 * int (*ndo_get_offload_stats)(int attr_id, const struct net_device *dev, 1131 * void *attr_data) 1132 * Get statistics for offload operations by attr_id. Write it into the 1133 * attr_data pointer. 1134 * 1135 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16 vid); 1136 * If device supports VLAN filtering this function is called when a 1137 * VLAN id is registered. 1138 * 1139 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, __be16 proto, u16 vid); 1140 * If device supports VLAN filtering this function is called when a 1141 * VLAN id is unregistered. 1142 * 1143 * void (*ndo_poll_controller)(struct net_device *dev); 1144 * 1145 * SR-IOV management functions. 1146 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac); 1147 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan, 1148 * u8 qos, __be16 proto); 1149 * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate, 1150 * int max_tx_rate); 1151 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting); 1152 * int (*ndo_set_vf_trust)(struct net_device *dev, int vf, bool setting); 1153 * int (*ndo_get_vf_config)(struct net_device *dev, 1154 * int vf, struct ifla_vf_info *ivf); 1155 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state); 1156 * int (*ndo_set_vf_port)(struct net_device *dev, int vf, 1157 * struct nlattr *port[]); 1158 * 1159 * Enable or disable the VF ability to query its RSS Redirection Table and 1160 * Hash Key. This is needed since on some devices VF share this information 1161 * with PF and querying it may introduce a theoretical security risk. 1162 * int (*ndo_set_vf_rss_query_en)(struct net_device *dev, int vf, bool setting); 1163 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb); 1164 * int (*ndo_setup_tc)(struct net_device *dev, enum tc_setup_type type, 1165 * void *type_data); 1166 * Called to setup any 'tc' scheduler, classifier or action on @dev. 1167 * This is always called from the stack with the rtnl lock held and netif 1168 * tx queues stopped. This allows the netdevice to perform queue 1169 * management safely. 1170 * 1171 * Fiber Channel over Ethernet (FCoE) offload functions. 1172 * int (*ndo_fcoe_enable)(struct net_device *dev); 1173 * Called when the FCoE protocol stack wants to start using LLD for FCoE 1174 * so the underlying device can perform whatever needed configuration or 1175 * initialization to support acceleration of FCoE traffic. 1176 * 1177 * int (*ndo_fcoe_disable)(struct net_device *dev); 1178 * Called when the FCoE protocol stack wants to stop using LLD for FCoE 1179 * so the underlying device can perform whatever needed clean-ups to 1180 * stop supporting acceleration of FCoE traffic. 1181 * 1182 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid, 1183 * struct scatterlist *sgl, unsigned int sgc); 1184 * Called when the FCoE Initiator wants to initialize an I/O that 1185 * is a possible candidate for Direct Data Placement (DDP). The LLD can 1186 * perform necessary setup and returns 1 to indicate the device is set up 1187 * successfully to perform DDP on this I/O, otherwise this returns 0. 1188 * 1189 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid); 1190 * Called when the FCoE Initiator/Target is done with the DDPed I/O as 1191 * indicated by the FC exchange id 'xid', so the underlying device can 1192 * clean up and reuse resources for later DDP requests. 1193 * 1194 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid, 1195 * struct scatterlist *sgl, unsigned int sgc); 1196 * Called when the FCoE Target wants to initialize an I/O that 1197 * is a possible candidate for Direct Data Placement (DDP). The LLD can 1198 * perform necessary setup and returns 1 to indicate the device is set up 1199 * successfully to perform DDP on this I/O, otherwise this returns 0. 1200 * 1201 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev, 1202 * struct netdev_fcoe_hbainfo *hbainfo); 1203 * Called when the FCoE Protocol stack wants information on the underlying 1204 * device. This information is utilized by the FCoE protocol stack to 1205 * register attributes with Fiber Channel management service as per the 1206 * FC-GS Fabric Device Management Information(FDMI) specification. 1207 * 1208 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type); 1209 * Called when the underlying device wants to override default World Wide 1210 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own 1211 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE 1212 * protocol stack to use. 1213 * 1214 * RFS acceleration. 1215 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb, 1216 * u16 rxq_index, u32 flow_id); 1217 * Set hardware filter for RFS. rxq_index is the target queue index; 1218 * flow_id is a flow ID to be passed to rps_may_expire_flow() later. 1219 * Return the filter ID on success, or a negative error code. 1220 * 1221 * Slave management functions (for bridge, bonding, etc). 1222 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev); 1223 * Called to make another netdev an underling. 1224 * 1225 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev); 1226 * Called to release previously enslaved netdev. 1227 * 1228 * struct net_device *(*ndo_get_xmit_slave)(struct net_device *dev, 1229 * struct sk_buff *skb, 1230 * bool all_slaves); 1231 * Get the xmit slave of master device. If all_slaves is true, function 1232 * assume all the slaves can transmit. 1233 * 1234 * Feature/offload setting functions. 1235 * netdev_features_t (*ndo_fix_features)(struct net_device *dev, 1236 * netdev_features_t features); 1237 * Adjusts the requested feature flags according to device-specific 1238 * constraints, and returns the resulting flags. Must not modify 1239 * the device state. 1240 * 1241 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features); 1242 * Called to update device configuration to new features. Passed 1243 * feature set might be less than what was returned by ndo_fix_features()). 1244 * Must return >0 or -errno if it changed dev->features itself. 1245 * 1246 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[], 1247 * struct net_device *dev, 1248 * const unsigned char *addr, u16 vid, u16 flags, 1249 * struct netlink_ext_ack *extack); 1250 * Adds an FDB entry to dev for addr. 1251 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[], 1252 * struct net_device *dev, 1253 * const unsigned char *addr, u16 vid) 1254 * Deletes the FDB entry from dev coresponding to addr. 1255 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb, 1256 * struct net_device *dev, struct net_device *filter_dev, 1257 * int *idx) 1258 * Used to add FDB entries to dump requests. Implementers should add 1259 * entries to skb and update idx with the number of entries. 1260 * 1261 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh, 1262 * u16 flags, struct netlink_ext_ack *extack) 1263 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq, 1264 * struct net_device *dev, u32 filter_mask, 1265 * int nlflags) 1266 * int (*ndo_bridge_dellink)(struct net_device *dev, struct nlmsghdr *nlh, 1267 * u16 flags); 1268 * 1269 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier); 1270 * Called to change device carrier. Soft-devices (like dummy, team, etc) 1271 * which do not represent real hardware may define this to allow their 1272 * userspace components to manage their virtual carrier state. Devices 1273 * that determine carrier state from physical hardware properties (eg 1274 * network cables) or protocol-dependent mechanisms (eg 1275 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function. 1276 * 1277 * int (*ndo_get_phys_port_id)(struct net_device *dev, 1278 * struct netdev_phys_item_id *ppid); 1279 * Called to get ID of physical port of this device. If driver does 1280 * not implement this, it is assumed that the hw is not able to have 1281 * multiple net devices on single physical port. 1282 * 1283 * int (*ndo_get_port_parent_id)(struct net_device *dev, 1284 * struct netdev_phys_item_id *ppid) 1285 * Called to get the parent ID of the physical port of this device. 1286 * 1287 * void* (*ndo_dfwd_add_station)(struct net_device *pdev, 1288 * struct net_device *dev) 1289 * Called by upper layer devices to accelerate switching or other 1290 * station functionality into hardware. 'pdev is the lowerdev 1291 * to use for the offload and 'dev' is the net device that will 1292 * back the offload. Returns a pointer to the private structure 1293 * the upper layer will maintain. 1294 * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv) 1295 * Called by upper layer device to delete the station created 1296 * by 'ndo_dfwd_add_station'. 'pdev' is the net device backing 1297 * the station and priv is the structure returned by the add 1298 * operation. 1299 * int (*ndo_set_tx_maxrate)(struct net_device *dev, 1300 * int queue_index, u32 maxrate); 1301 * Called when a user wants to set a max-rate limitation of specific 1302 * TX queue. 1303 * int (*ndo_get_iflink)(const struct net_device *dev); 1304 * Called to get the iflink value of this device. 1305 * int (*ndo_fill_metadata_dst)(struct net_device *dev, struct sk_buff *skb); 1306 * This function is used to get egress tunnel information for given skb. 1307 * This is useful for retrieving outer tunnel header parameters while 1308 * sampling packet. 1309 * void (*ndo_set_rx_headroom)(struct net_device *dev, int needed_headroom); 1310 * This function is used to specify the headroom that the skb must 1311 * consider when allocation skb during packet reception. Setting 1312 * appropriate rx headroom value allows avoiding skb head copy on 1313 * forward. Setting a negative value resets the rx headroom to the 1314 * default value. 1315 * int (*ndo_bpf)(struct net_device *dev, struct netdev_bpf *bpf); 1316 * This function is used to set or query state related to XDP on the 1317 * netdevice and manage BPF offload. See definition of 1318 * enum bpf_netdev_command for details. 1319 * int (*ndo_xdp_xmit)(struct net_device *dev, int n, struct xdp_frame **xdp, 1320 * u32 flags); 1321 * This function is used to submit @n XDP packets for transmit on a 1322 * netdevice. Returns number of frames successfully transmitted, frames 1323 * that got dropped are freed/returned via xdp_return_frame(). 1324 * Returns negative number, means general error invoking ndo, meaning 1325 * no frames were xmit'ed and core-caller will free all frames. 1326 * struct net_device *(*ndo_xdp_get_xmit_slave)(struct net_device *dev, 1327 * struct xdp_buff *xdp); 1328 * Get the xmit slave of master device based on the xdp_buff. 1329 * int (*ndo_xsk_wakeup)(struct net_device *dev, u32 queue_id, u32 flags); 1330 * This function is used to wake up the softirq, ksoftirqd or kthread 1331 * responsible for sending and/or receiving packets on a specific 1332 * queue id bound to an AF_XDP socket. The flags field specifies if 1333 * only RX, only Tx, or both should be woken up using the flags 1334 * XDP_WAKEUP_RX and XDP_WAKEUP_TX. 1335 * struct devlink_port *(*ndo_get_devlink_port)(struct net_device *dev); 1336 * Get devlink port instance associated with a given netdev. 1337 * Called with a reference on the netdevice and devlink locks only, 1338 * rtnl_lock is not held. 1339 * int (*ndo_tunnel_ctl)(struct net_device *dev, struct ip_tunnel_parm *p, 1340 * int cmd); 1341 * Add, change, delete or get information on an IPv4 tunnel. 1342 * struct net_device *(*ndo_get_peer_dev)(struct net_device *dev); 1343 * If a device is paired with a peer device, return the peer instance. 1344 * The caller must be under RCU read context. 1345 * int (*ndo_fill_forward_path)(struct net_device_path_ctx *ctx, struct net_device_path *path); 1346 * Get the forwarding path to reach the real device from the HW destination address 1347 */ 1348 struct net_device_ops { 1349 int (*ndo_init)(struct net_device *dev); 1350 void (*ndo_uninit)(struct net_device *dev); 1351 int (*ndo_open)(struct net_device *dev); 1352 int (*ndo_stop)(struct net_device *dev); 1353 netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb, 1354 struct net_device *dev); 1355 netdev_features_t (*ndo_features_check)(struct sk_buff *skb, 1356 struct net_device *dev, 1357 netdev_features_t features); 1358 u16 (*ndo_select_queue)(struct net_device *dev, 1359 struct sk_buff *skb, 1360 struct net_device *sb_dev); 1361 void (*ndo_change_rx_flags)(struct net_device *dev, 1362 int flags); 1363 void (*ndo_set_rx_mode)(struct net_device *dev); 1364 int (*ndo_set_mac_address)(struct net_device *dev, 1365 void *addr); 1366 int (*ndo_validate_addr)(struct net_device *dev); 1367 int (*ndo_do_ioctl)(struct net_device *dev, 1368 struct ifreq *ifr, int cmd); 1369 int (*ndo_eth_ioctl)(struct net_device *dev, 1370 struct ifreq *ifr, int cmd); 1371 int (*ndo_siocbond)(struct net_device *dev, 1372 struct ifreq *ifr, int cmd); 1373 int (*ndo_siocwandev)(struct net_device *dev, 1374 struct if_settings *ifs); 1375 int (*ndo_siocdevprivate)(struct net_device *dev, 1376 struct ifreq *ifr, 1377 void __user *data, int cmd); 1378 int (*ndo_set_config)(struct net_device *dev, 1379 struct ifmap *map); 1380 int (*ndo_change_mtu)(struct net_device *dev, 1381 int new_mtu); 1382 int (*ndo_neigh_setup)(struct net_device *dev, 1383 struct neigh_parms *); 1384 void (*ndo_tx_timeout) (struct net_device *dev, 1385 unsigned int txqueue); 1386 1387 void (*ndo_get_stats64)(struct net_device *dev, 1388 struct rtnl_link_stats64 *storage); 1389 bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id); 1390 int (*ndo_get_offload_stats)(int attr_id, 1391 const struct net_device *dev, 1392 void *attr_data); 1393 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev); 1394 1395 int (*ndo_vlan_rx_add_vid)(struct net_device *dev, 1396 __be16 proto, u16 vid); 1397 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, 1398 __be16 proto, u16 vid); 1399 #ifdef CONFIG_NET_POLL_CONTROLLER 1400 void (*ndo_poll_controller)(struct net_device *dev); 1401 int (*ndo_netpoll_setup)(struct net_device *dev, 1402 struct netpoll_info *info); 1403 void (*ndo_netpoll_cleanup)(struct net_device *dev); 1404 #endif 1405 int (*ndo_set_vf_mac)(struct net_device *dev, 1406 int queue, u8 *mac); 1407 int (*ndo_set_vf_vlan)(struct net_device *dev, 1408 int queue, u16 vlan, 1409 u8 qos, __be16 proto); 1410 int (*ndo_set_vf_rate)(struct net_device *dev, 1411 int vf, int min_tx_rate, 1412 int max_tx_rate); 1413 int (*ndo_set_vf_spoofchk)(struct net_device *dev, 1414 int vf, bool setting); 1415 int (*ndo_set_vf_trust)(struct net_device *dev, 1416 int vf, bool setting); 1417 int (*ndo_get_vf_config)(struct net_device *dev, 1418 int vf, 1419 struct ifla_vf_info *ivf); 1420 int (*ndo_set_vf_link_state)(struct net_device *dev, 1421 int vf, int link_state); 1422 int (*ndo_get_vf_stats)(struct net_device *dev, 1423 int vf, 1424 struct ifla_vf_stats 1425 *vf_stats); 1426 int (*ndo_set_vf_port)(struct net_device *dev, 1427 int vf, 1428 struct nlattr *port[]); 1429 int (*ndo_get_vf_port)(struct net_device *dev, 1430 int vf, struct sk_buff *skb); 1431 int (*ndo_get_vf_guid)(struct net_device *dev, 1432 int vf, 1433 struct ifla_vf_guid *node_guid, 1434 struct ifla_vf_guid *port_guid); 1435 int (*ndo_set_vf_guid)(struct net_device *dev, 1436 int vf, u64 guid, 1437 int guid_type); 1438 int (*ndo_set_vf_rss_query_en)( 1439 struct net_device *dev, 1440 int vf, bool setting); 1441 int (*ndo_setup_tc)(struct net_device *dev, 1442 enum tc_setup_type type, 1443 void *type_data); 1444 #if IS_ENABLED(CONFIG_FCOE) 1445 int (*ndo_fcoe_enable)(struct net_device *dev); 1446 int (*ndo_fcoe_disable)(struct net_device *dev); 1447 int (*ndo_fcoe_ddp_setup)(struct net_device *dev, 1448 u16 xid, 1449 struct scatterlist *sgl, 1450 unsigned int sgc); 1451 int (*ndo_fcoe_ddp_done)(struct net_device *dev, 1452 u16 xid); 1453 int (*ndo_fcoe_ddp_target)(struct net_device *dev, 1454 u16 xid, 1455 struct scatterlist *sgl, 1456 unsigned int sgc); 1457 int (*ndo_fcoe_get_hbainfo)(struct net_device *dev, 1458 struct netdev_fcoe_hbainfo *hbainfo); 1459 #endif 1460 1461 #if IS_ENABLED(CONFIG_LIBFCOE) 1462 #define NETDEV_FCOE_WWNN 0 1463 #define NETDEV_FCOE_WWPN 1 1464 int (*ndo_fcoe_get_wwn)(struct net_device *dev, 1465 u64 *wwn, int type); 1466 #endif 1467 1468 #ifdef CONFIG_RFS_ACCEL 1469 int (*ndo_rx_flow_steer)(struct net_device *dev, 1470 const struct sk_buff *skb, 1471 u16 rxq_index, 1472 u32 flow_id); 1473 #endif 1474 int (*ndo_add_slave)(struct net_device *dev, 1475 struct net_device *slave_dev, 1476 struct netlink_ext_ack *extack); 1477 int (*ndo_del_slave)(struct net_device *dev, 1478 struct net_device *slave_dev); 1479 struct net_device* (*ndo_get_xmit_slave)(struct net_device *dev, 1480 struct sk_buff *skb, 1481 bool all_slaves); 1482 struct net_device* (*ndo_sk_get_lower_dev)(struct net_device *dev, 1483 struct sock *sk); 1484 netdev_features_t (*ndo_fix_features)(struct net_device *dev, 1485 netdev_features_t features); 1486 int (*ndo_set_features)(struct net_device *dev, 1487 netdev_features_t features); 1488 int (*ndo_neigh_construct)(struct net_device *dev, 1489 struct neighbour *n); 1490 void (*ndo_neigh_destroy)(struct net_device *dev, 1491 struct neighbour *n); 1492 1493 int (*ndo_fdb_add)(struct ndmsg *ndm, 1494 struct nlattr *tb[], 1495 struct net_device *dev, 1496 const unsigned char *addr, 1497 u16 vid, 1498 u16 flags, 1499 struct netlink_ext_ack *extack); 1500 int (*ndo_fdb_del)(struct ndmsg *ndm, 1501 struct nlattr *tb[], 1502 struct net_device *dev, 1503 const unsigned char *addr, 1504 u16 vid); 1505 int (*ndo_fdb_dump)(struct sk_buff *skb, 1506 struct netlink_callback *cb, 1507 struct net_device *dev, 1508 struct net_device *filter_dev, 1509 int *idx); 1510 int (*ndo_fdb_get)(struct sk_buff *skb, 1511 struct nlattr *tb[], 1512 struct net_device *dev, 1513 const unsigned char *addr, 1514 u16 vid, u32 portid, u32 seq, 1515 struct netlink_ext_ack *extack); 1516 int (*ndo_bridge_setlink)(struct net_device *dev, 1517 struct nlmsghdr *nlh, 1518 u16 flags, 1519 struct netlink_ext_ack *extack); 1520 int (*ndo_bridge_getlink)(struct sk_buff *skb, 1521 u32 pid, u32 seq, 1522 struct net_device *dev, 1523 u32 filter_mask, 1524 int nlflags); 1525 int (*ndo_bridge_dellink)(struct net_device *dev, 1526 struct nlmsghdr *nlh, 1527 u16 flags); 1528 int (*ndo_change_carrier)(struct net_device *dev, 1529 bool new_carrier); 1530 int (*ndo_get_phys_port_id)(struct net_device *dev, 1531 struct netdev_phys_item_id *ppid); 1532 int (*ndo_get_port_parent_id)(struct net_device *dev, 1533 struct netdev_phys_item_id *ppid); 1534 int (*ndo_get_phys_port_name)(struct net_device *dev, 1535 char *name, size_t len); 1536 void* (*ndo_dfwd_add_station)(struct net_device *pdev, 1537 struct net_device *dev); 1538 void (*ndo_dfwd_del_station)(struct net_device *pdev, 1539 void *priv); 1540 1541 int (*ndo_set_tx_maxrate)(struct net_device *dev, 1542 int queue_index, 1543 u32 maxrate); 1544 int (*ndo_get_iflink)(const struct net_device *dev); 1545 int (*ndo_fill_metadata_dst)(struct net_device *dev, 1546 struct sk_buff *skb); 1547 void (*ndo_set_rx_headroom)(struct net_device *dev, 1548 int needed_headroom); 1549 int (*ndo_bpf)(struct net_device *dev, 1550 struct netdev_bpf *bpf); 1551 int (*ndo_xdp_xmit)(struct net_device *dev, int n, 1552 struct xdp_frame **xdp, 1553 u32 flags); 1554 struct net_device * (*ndo_xdp_get_xmit_slave)(struct net_device *dev, 1555 struct xdp_buff *xdp); 1556 int (*ndo_xsk_wakeup)(struct net_device *dev, 1557 u32 queue_id, u32 flags); 1558 struct devlink_port * (*ndo_get_devlink_port)(struct net_device *dev); 1559 int (*ndo_tunnel_ctl)(struct net_device *dev, 1560 struct ip_tunnel_parm *p, int cmd); 1561 struct net_device * (*ndo_get_peer_dev)(struct net_device *dev); 1562 int (*ndo_fill_forward_path)(struct net_device_path_ctx *ctx, 1563 struct net_device_path *path); 1564 }; 1565 1566 /** 1567 * enum netdev_priv_flags - &struct net_device priv_flags 1568 * 1569 * These are the &struct net_device, they are only set internally 1570 * by drivers and used in the kernel. These flags are invisible to 1571 * userspace; this means that the order of these flags can change 1572 * during any kernel release. 1573 * 1574 * You should have a pretty good reason to be extending these flags. 1575 * 1576 * @IFF_802_1Q_VLAN: 802.1Q VLAN device 1577 * @IFF_EBRIDGE: Ethernet bridging device 1578 * @IFF_BONDING: bonding master or slave 1579 * @IFF_ISATAP: ISATAP interface (RFC4214) 1580 * @IFF_WAN_HDLC: WAN HDLC device 1581 * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to 1582 * release skb->dst 1583 * @IFF_DONT_BRIDGE: disallow bridging this ether dev 1584 * @IFF_DISABLE_NETPOLL: disable netpoll at run-time 1585 * @IFF_MACVLAN_PORT: device used as macvlan port 1586 * @IFF_BRIDGE_PORT: device used as bridge port 1587 * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port 1588 * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit 1589 * @IFF_UNICAST_FLT: Supports unicast filtering 1590 * @IFF_TEAM_PORT: device used as team port 1591 * @IFF_SUPP_NOFCS: device supports sending custom FCS 1592 * @IFF_LIVE_ADDR_CHANGE: device supports hardware address 1593 * change when it's running 1594 * @IFF_MACVLAN: Macvlan device 1595 * @IFF_XMIT_DST_RELEASE_PERM: IFF_XMIT_DST_RELEASE not taking into account 1596 * underlying stacked devices 1597 * @IFF_L3MDEV_MASTER: device is an L3 master device 1598 * @IFF_NO_QUEUE: device can run without qdisc attached 1599 * @IFF_OPENVSWITCH: device is a Open vSwitch master 1600 * @IFF_L3MDEV_SLAVE: device is enslaved to an L3 master device 1601 * @IFF_TEAM: device is a team device 1602 * @IFF_RXFH_CONFIGURED: device has had Rx Flow indirection table configured 1603 * @IFF_PHONY_HEADROOM: the headroom value is controlled by an external 1604 * entity (i.e. the master device for bridged veth) 1605 * @IFF_MACSEC: device is a MACsec device 1606 * @IFF_NO_RX_HANDLER: device doesn't support the rx_handler hook 1607 * @IFF_FAILOVER: device is a failover master device 1608 * @IFF_FAILOVER_SLAVE: device is lower dev of a failover master device 1609 * @IFF_L3MDEV_RX_HANDLER: only invoke the rx handler of L3 master device 1610 * @IFF_LIVE_RENAME_OK: rename is allowed while device is up and running 1611 * @IFF_TX_SKB_NO_LINEAR: device/driver is capable of xmitting frames with 1612 * skb_headlen(skb) == 0 (data starts from frag0) 1613 * @IFF_CHANGE_PROTO_DOWN: device supports setting carrier via IFLA_PROTO_DOWN 1614 */ 1615 enum netdev_priv_flags { 1616 IFF_802_1Q_VLAN = 1<<0, 1617 IFF_EBRIDGE = 1<<1, 1618 IFF_BONDING = 1<<2, 1619 IFF_ISATAP = 1<<3, 1620 IFF_WAN_HDLC = 1<<4, 1621 IFF_XMIT_DST_RELEASE = 1<<5, 1622 IFF_DONT_BRIDGE = 1<<6, 1623 IFF_DISABLE_NETPOLL = 1<<7, 1624 IFF_MACVLAN_PORT = 1<<8, 1625 IFF_BRIDGE_PORT = 1<<9, 1626 IFF_OVS_DATAPATH = 1<<10, 1627 IFF_TX_SKB_SHARING = 1<<11, 1628 IFF_UNICAST_FLT = 1<<12, 1629 IFF_TEAM_PORT = 1<<13, 1630 IFF_SUPP_NOFCS = 1<<14, 1631 IFF_LIVE_ADDR_CHANGE = 1<<15, 1632 IFF_MACVLAN = 1<<16, 1633 IFF_XMIT_DST_RELEASE_PERM = 1<<17, 1634 IFF_L3MDEV_MASTER = 1<<18, 1635 IFF_NO_QUEUE = 1<<19, 1636 IFF_OPENVSWITCH = 1<<20, 1637 IFF_L3MDEV_SLAVE = 1<<21, 1638 IFF_TEAM = 1<<22, 1639 IFF_RXFH_CONFIGURED = 1<<23, 1640 IFF_PHONY_HEADROOM = 1<<24, 1641 IFF_MACSEC = 1<<25, 1642 IFF_NO_RX_HANDLER = 1<<26, 1643 IFF_FAILOVER = 1<<27, 1644 IFF_FAILOVER_SLAVE = 1<<28, 1645 IFF_L3MDEV_RX_HANDLER = 1<<29, 1646 IFF_LIVE_RENAME_OK = 1<<30, 1647 IFF_TX_SKB_NO_LINEAR = 1<<31, 1648 IFF_CHANGE_PROTO_DOWN = BIT_ULL(32), 1649 }; 1650 1651 #define IFF_802_1Q_VLAN IFF_802_1Q_VLAN 1652 #define IFF_EBRIDGE IFF_EBRIDGE 1653 #define IFF_BONDING IFF_BONDING 1654 #define IFF_ISATAP IFF_ISATAP 1655 #define IFF_WAN_HDLC IFF_WAN_HDLC 1656 #define IFF_XMIT_DST_RELEASE IFF_XMIT_DST_RELEASE 1657 #define IFF_DONT_BRIDGE IFF_DONT_BRIDGE 1658 #define IFF_DISABLE_NETPOLL IFF_DISABLE_NETPOLL 1659 #define IFF_MACVLAN_PORT IFF_MACVLAN_PORT 1660 #define IFF_BRIDGE_PORT IFF_BRIDGE_PORT 1661 #define IFF_OVS_DATAPATH IFF_OVS_DATAPATH 1662 #define IFF_TX_SKB_SHARING IFF_TX_SKB_SHARING 1663 #define IFF_UNICAST_FLT IFF_UNICAST_FLT 1664 #define IFF_TEAM_PORT IFF_TEAM_PORT 1665 #define IFF_SUPP_NOFCS IFF_SUPP_NOFCS 1666 #define IFF_LIVE_ADDR_CHANGE IFF_LIVE_ADDR_CHANGE 1667 #define IFF_MACVLAN IFF_MACVLAN 1668 #define IFF_XMIT_DST_RELEASE_PERM IFF_XMIT_DST_RELEASE_PERM 1669 #define IFF_L3MDEV_MASTER IFF_L3MDEV_MASTER 1670 #define IFF_NO_QUEUE IFF_NO_QUEUE 1671 #define IFF_OPENVSWITCH IFF_OPENVSWITCH 1672 #define IFF_L3MDEV_SLAVE IFF_L3MDEV_SLAVE 1673 #define IFF_TEAM IFF_TEAM 1674 #define IFF_RXFH_CONFIGURED IFF_RXFH_CONFIGURED 1675 #define IFF_PHONY_HEADROOM IFF_PHONY_HEADROOM 1676 #define IFF_MACSEC IFF_MACSEC 1677 #define IFF_NO_RX_HANDLER IFF_NO_RX_HANDLER 1678 #define IFF_FAILOVER IFF_FAILOVER 1679 #define IFF_FAILOVER_SLAVE IFF_FAILOVER_SLAVE 1680 #define IFF_L3MDEV_RX_HANDLER IFF_L3MDEV_RX_HANDLER 1681 #define IFF_LIVE_RENAME_OK IFF_LIVE_RENAME_OK 1682 #define IFF_TX_SKB_NO_LINEAR IFF_TX_SKB_NO_LINEAR 1683 1684 /* Specifies the type of the struct net_device::ml_priv pointer */ 1685 enum netdev_ml_priv_type { 1686 ML_PRIV_NONE, 1687 ML_PRIV_CAN, 1688 }; 1689 1690 /** 1691 * struct net_device - The DEVICE structure. 1692 * 1693 * Actually, this whole structure is a big mistake. It mixes I/O 1694 * data with strictly "high-level" data, and it has to know about 1695 * almost every data structure used in the INET module. 1696 * 1697 * @name: This is the first field of the "visible" part of this structure 1698 * (i.e. as seen by users in the "Space.c" file). It is the name 1699 * of the interface. 1700 * 1701 * @name_node: Name hashlist node 1702 * @ifalias: SNMP alias 1703 * @mem_end: Shared memory end 1704 * @mem_start: Shared memory start 1705 * @base_addr: Device I/O address 1706 * @irq: Device IRQ number 1707 * 1708 * @state: Generic network queuing layer state, see netdev_state_t 1709 * @dev_list: The global list of network devices 1710 * @napi_list: List entry used for polling NAPI devices 1711 * @unreg_list: List entry when we are unregistering the 1712 * device; see the function unregister_netdev 1713 * @close_list: List entry used when we are closing the device 1714 * @ptype_all: Device-specific packet handlers for all protocols 1715 * @ptype_specific: Device-specific, protocol-specific packet handlers 1716 * 1717 * @adj_list: Directly linked devices, like slaves for bonding 1718 * @features: Currently active device features 1719 * @hw_features: User-changeable features 1720 * 1721 * @wanted_features: User-requested features 1722 * @vlan_features: Mask of features inheritable by VLAN devices 1723 * 1724 * @hw_enc_features: Mask of features inherited by encapsulating devices 1725 * This field indicates what encapsulation 1726 * offloads the hardware is capable of doing, 1727 * and drivers will need to set them appropriately. 1728 * 1729 * @mpls_features: Mask of features inheritable by MPLS 1730 * @gso_partial_features: value(s) from NETIF_F_GSO\* 1731 * 1732 * @ifindex: interface index 1733 * @group: The group the device belongs to 1734 * 1735 * @stats: Statistics struct, which was left as a legacy, use 1736 * rtnl_link_stats64 instead 1737 * 1738 * @rx_dropped: Dropped packets by core network, 1739 * do not use this in drivers 1740 * @tx_dropped: Dropped packets by core network, 1741 * do not use this in drivers 1742 * @rx_nohandler: nohandler dropped packets by core network on 1743 * inactive devices, do not use this in drivers 1744 * @carrier_up_count: Number of times the carrier has been up 1745 * @carrier_down_count: Number of times the carrier has been down 1746 * 1747 * @wireless_handlers: List of functions to handle Wireless Extensions, 1748 * instead of ioctl, 1749 * see <net/iw_handler.h> for details. 1750 * @wireless_data: Instance data managed by the core of wireless extensions 1751 * 1752 * @netdev_ops: Includes several pointers to callbacks, 1753 * if one wants to override the ndo_*() functions 1754 * @ethtool_ops: Management operations 1755 * @l3mdev_ops: Layer 3 master device operations 1756 * @ndisc_ops: Includes callbacks for different IPv6 neighbour 1757 * discovery handling. Necessary for e.g. 6LoWPAN. 1758 * @xfrmdev_ops: Transformation offload operations 1759 * @tlsdev_ops: Transport Layer Security offload operations 1760 * @header_ops: Includes callbacks for creating,parsing,caching,etc 1761 * of Layer 2 headers. 1762 * 1763 * @flags: Interface flags (a la BSD) 1764 * @priv_flags: Like 'flags' but invisible to userspace, 1765 * see if.h for the definitions 1766 * @gflags: Global flags ( kept as legacy ) 1767 * @padded: How much padding added by alloc_netdev() 1768 * @operstate: RFC2863 operstate 1769 * @link_mode: Mapping policy to operstate 1770 * @if_port: Selectable AUI, TP, ... 1771 * @dma: DMA channel 1772 * @mtu: Interface MTU value 1773 * @min_mtu: Interface Minimum MTU value 1774 * @max_mtu: Interface Maximum MTU value 1775 * @type: Interface hardware type 1776 * @hard_header_len: Maximum hardware header length. 1777 * @min_header_len: Minimum hardware header length 1778 * 1779 * @needed_headroom: Extra headroom the hardware may need, but not in all 1780 * cases can this be guaranteed 1781 * @needed_tailroom: Extra tailroom the hardware may need, but not in all 1782 * cases can this be guaranteed. Some cases also use 1783 * LL_MAX_HEADER instead to allocate the skb 1784 * 1785 * interface address info: 1786 * 1787 * @perm_addr: Permanent hw address 1788 * @addr_assign_type: Hw address assignment type 1789 * @addr_len: Hardware address length 1790 * @upper_level: Maximum depth level of upper devices. 1791 * @lower_level: Maximum depth level of lower devices. 1792 * @neigh_priv_len: Used in neigh_alloc() 1793 * @dev_id: Used to differentiate devices that share 1794 * the same link layer address 1795 * @dev_port: Used to differentiate devices that share 1796 * the same function 1797 * @addr_list_lock: XXX: need comments on this one 1798 * @name_assign_type: network interface name assignment type 1799 * @uc_promisc: Counter that indicates promiscuous mode 1800 * has been enabled due to the need to listen to 1801 * additional unicast addresses in a device that 1802 * does not implement ndo_set_rx_mode() 1803 * @uc: unicast mac addresses 1804 * @mc: multicast mac addresses 1805 * @dev_addrs: list of device hw addresses 1806 * @queues_kset: Group of all Kobjects in the Tx and RX queues 1807 * @promiscuity: Number of times the NIC is told to work in 1808 * promiscuous mode; if it becomes 0 the NIC will 1809 * exit promiscuous mode 1810 * @allmulti: Counter, enables or disables allmulticast mode 1811 * 1812 * @vlan_info: VLAN info 1813 * @dsa_ptr: dsa specific data 1814 * @tipc_ptr: TIPC specific data 1815 * @atalk_ptr: AppleTalk link 1816 * @ip_ptr: IPv4 specific data 1817 * @dn_ptr: DECnet specific data 1818 * @ip6_ptr: IPv6 specific data 1819 * @ax25_ptr: AX.25 specific data 1820 * @ieee80211_ptr: IEEE 802.11 specific data, assign before registering 1821 * @ieee802154_ptr: IEEE 802.15.4 low-rate Wireless Personal Area Network 1822 * device struct 1823 * @mpls_ptr: mpls_dev struct pointer 1824 * @mctp_ptr: MCTP specific data 1825 * 1826 * @dev_addr: Hw address (before bcast, 1827 * because most packets are unicast) 1828 * 1829 * @_rx: Array of RX queues 1830 * @num_rx_queues: Number of RX queues 1831 * allocated at register_netdev() time 1832 * @real_num_rx_queues: Number of RX queues currently active in device 1833 * @xdp_prog: XDP sockets filter program pointer 1834 * @gro_flush_timeout: timeout for GRO layer in NAPI 1835 * @napi_defer_hard_irqs: If not zero, provides a counter that would 1836 * allow to avoid NIC hard IRQ, on busy queues. 1837 * 1838 * @rx_handler: handler for received packets 1839 * @rx_handler_data: XXX: need comments on this one 1840 * @miniq_ingress: ingress/clsact qdisc specific data for 1841 * ingress processing 1842 * @ingress_queue: XXX: need comments on this one 1843 * @nf_hooks_ingress: netfilter hooks executed for ingress packets 1844 * @broadcast: hw bcast address 1845 * 1846 * @rx_cpu_rmap: CPU reverse-mapping for RX completion interrupts, 1847 * indexed by RX queue number. Assigned by driver. 1848 * This must only be set if the ndo_rx_flow_steer 1849 * operation is defined 1850 * @index_hlist: Device index hash chain 1851 * 1852 * @_tx: Array of TX queues 1853 * @num_tx_queues: Number of TX queues allocated at alloc_netdev_mq() time 1854 * @real_num_tx_queues: Number of TX queues currently active in device 1855 * @qdisc: Root qdisc from userspace point of view 1856 * @tx_queue_len: Max frames per queue allowed 1857 * @tx_global_lock: XXX: need comments on this one 1858 * @xdp_bulkq: XDP device bulk queue 1859 * @xps_maps: all CPUs/RXQs maps for XPS device 1860 * 1861 * @xps_maps: XXX: need comments on this one 1862 * @miniq_egress: clsact qdisc specific data for 1863 * egress processing 1864 * @nf_hooks_egress: netfilter hooks executed for egress packets 1865 * @qdisc_hash: qdisc hash table 1866 * @watchdog_timeo: Represents the timeout that is used by 1867 * the watchdog (see dev_watchdog()) 1868 * @watchdog_timer: List of timers 1869 * 1870 * @proto_down_reason: reason a netdev interface is held down 1871 * @pcpu_refcnt: Number of references to this device 1872 * @dev_refcnt: Number of references to this device 1873 * @refcnt_tracker: Tracker directory for tracked references to this device 1874 * @todo_list: Delayed register/unregister 1875 * @link_watch_list: XXX: need comments on this one 1876 * 1877 * @reg_state: Register/unregister state machine 1878 * @dismantle: Device is going to be freed 1879 * @rtnl_link_state: This enum represents the phases of creating 1880 * a new link 1881 * 1882 * @needs_free_netdev: Should unregister perform free_netdev? 1883 * @priv_destructor: Called from unregister 1884 * @npinfo: XXX: need comments on this one 1885 * @nd_net: Network namespace this network device is inside 1886 * 1887 * @ml_priv: Mid-layer private 1888 * @ml_priv_type: Mid-layer private type 1889 * @lstats: Loopback statistics 1890 * @tstats: Tunnel statistics 1891 * @dstats: Dummy statistics 1892 * @vstats: Virtual ethernet statistics 1893 * 1894 * @garp_port: GARP 1895 * @mrp_port: MRP 1896 * 1897 * @dev: Class/net/name entry 1898 * @sysfs_groups: Space for optional device, statistics and wireless 1899 * sysfs groups 1900 * 1901 * @sysfs_rx_queue_group: Space for optional per-rx queue attributes 1902 * @rtnl_link_ops: Rtnl_link_ops 1903 * 1904 * @gso_max_size: Maximum size of generic segmentation offload 1905 * @gso_max_segs: Maximum number of segments that can be passed to the 1906 * NIC for GSO 1907 * 1908 * @dcbnl_ops: Data Center Bridging netlink ops 1909 * @num_tc: Number of traffic classes in the net device 1910 * @tc_to_txq: XXX: need comments on this one 1911 * @prio_tc_map: XXX: need comments on this one 1912 * 1913 * @fcoe_ddp_xid: Max exchange id for FCoE LRO by ddp 1914 * 1915 * @priomap: XXX: need comments on this one 1916 * @phydev: Physical device may attach itself 1917 * for hardware timestamping 1918 * @sfp_bus: attached &struct sfp_bus structure. 1919 * 1920 * @qdisc_tx_busylock: lockdep class annotating Qdisc->busylock spinlock 1921 * 1922 * @proto_down: protocol port state information can be sent to the 1923 * switch driver and used to set the phys state of the 1924 * switch port. 1925 * 1926 * @wol_enabled: Wake-on-LAN is enabled 1927 * 1928 * @threaded: napi threaded mode is enabled 1929 * 1930 * @net_notifier_list: List of per-net netdev notifier block 1931 * that follow this device when it is moved 1932 * to another network namespace. 1933 * 1934 * @macsec_ops: MACsec offloading ops 1935 * 1936 * @udp_tunnel_nic_info: static structure describing the UDP tunnel 1937 * offload capabilities of the device 1938 * @udp_tunnel_nic: UDP tunnel offload state 1939 * @xdp_state: stores info on attached XDP BPF programs 1940 * 1941 * @nested_level: Used as a parameter of spin_lock_nested() of 1942 * dev->addr_list_lock. 1943 * @unlink_list: As netif_addr_lock() can be called recursively, 1944 * keep a list of interfaces to be deleted. 1945 * @gro_max_size: Maximum size of aggregated packet in generic 1946 * receive offload (GRO) 1947 * 1948 * @dev_addr_shadow: Copy of @dev_addr to catch direct writes. 1949 * @linkwatch_dev_tracker: refcount tracker used by linkwatch. 1950 * @watchdog_dev_tracker: refcount tracker used by watchdog. 1951 * @dev_registered_tracker: tracker for reference held while 1952 * registered 1953 * 1954 * FIXME: cleanup struct net_device such that network protocol info 1955 * moves out. 1956 */ 1957 1958 struct net_device { 1959 char name[IFNAMSIZ]; 1960 struct netdev_name_node *name_node; 1961 struct dev_ifalias __rcu *ifalias; 1962 /* 1963 * I/O specific fields 1964 * FIXME: Merge these and struct ifmap into one 1965 */ 1966 unsigned long mem_end; 1967 unsigned long mem_start; 1968 unsigned long base_addr; 1969 1970 /* 1971 * Some hardware also needs these fields (state,dev_list, 1972 * napi_list,unreg_list,close_list) but they are not 1973 * part of the usual set specified in Space.c. 1974 */ 1975 1976 unsigned long state; 1977 1978 struct list_head dev_list; 1979 struct list_head napi_list; 1980 struct list_head unreg_list; 1981 struct list_head close_list; 1982 struct list_head ptype_all; 1983 struct list_head ptype_specific; 1984 1985 struct { 1986 struct list_head upper; 1987 struct list_head lower; 1988 } adj_list; 1989 1990 /* Read-mostly cache-line for fast-path access */ 1991 unsigned int flags; 1992 unsigned long long priv_flags; 1993 const struct net_device_ops *netdev_ops; 1994 int ifindex; 1995 unsigned short gflags; 1996 unsigned short hard_header_len; 1997 1998 /* Note : dev->mtu is often read without holding a lock. 1999 * Writers usually hold RTNL. 2000 * It is recommended to use READ_ONCE() to annotate the reads, 2001 * and to use WRITE_ONCE() to annotate the writes. 2002 */ 2003 unsigned int mtu; 2004 unsigned short needed_headroom; 2005 unsigned short needed_tailroom; 2006 2007 netdev_features_t features; 2008 netdev_features_t hw_features; 2009 netdev_features_t wanted_features; 2010 netdev_features_t vlan_features; 2011 netdev_features_t hw_enc_features; 2012 netdev_features_t mpls_features; 2013 netdev_features_t gso_partial_features; 2014 2015 unsigned int min_mtu; 2016 unsigned int max_mtu; 2017 unsigned short type; 2018 unsigned char min_header_len; 2019 unsigned char name_assign_type; 2020 2021 int group; 2022 2023 struct net_device_stats stats; /* not used by modern drivers */ 2024 2025 atomic_long_t rx_dropped; 2026 atomic_long_t tx_dropped; 2027 atomic_long_t rx_nohandler; 2028 2029 /* Stats to monitor link on/off, flapping */ 2030 atomic_t carrier_up_count; 2031 atomic_t carrier_down_count; 2032 2033 #ifdef CONFIG_WIRELESS_EXT 2034 const struct iw_handler_def *wireless_handlers; 2035 struct iw_public_data *wireless_data; 2036 #endif 2037 const struct ethtool_ops *ethtool_ops; 2038 #ifdef CONFIG_NET_L3_MASTER_DEV 2039 const struct l3mdev_ops *l3mdev_ops; 2040 #endif 2041 #if IS_ENABLED(CONFIG_IPV6) 2042 const struct ndisc_ops *ndisc_ops; 2043 #endif 2044 2045 #ifdef CONFIG_XFRM_OFFLOAD 2046 const struct xfrmdev_ops *xfrmdev_ops; 2047 #endif 2048 2049 #if IS_ENABLED(CONFIG_TLS_DEVICE) 2050 const struct tlsdev_ops *tlsdev_ops; 2051 #endif 2052 2053 const struct header_ops *header_ops; 2054 2055 unsigned char operstate; 2056 unsigned char link_mode; 2057 2058 unsigned char if_port; 2059 unsigned char dma; 2060 2061 /* Interface address info. */ 2062 unsigned char perm_addr[MAX_ADDR_LEN]; 2063 unsigned char addr_assign_type; 2064 unsigned char addr_len; 2065 unsigned char upper_level; 2066 unsigned char lower_level; 2067 2068 unsigned short neigh_priv_len; 2069 unsigned short dev_id; 2070 unsigned short dev_port; 2071 unsigned short padded; 2072 2073 spinlock_t addr_list_lock; 2074 int irq; 2075 2076 struct netdev_hw_addr_list uc; 2077 struct netdev_hw_addr_list mc; 2078 struct netdev_hw_addr_list dev_addrs; 2079 2080 #ifdef CONFIG_SYSFS 2081 struct kset *queues_kset; 2082 #endif 2083 #ifdef CONFIG_LOCKDEP 2084 struct list_head unlink_list; 2085 #endif 2086 unsigned int promiscuity; 2087 unsigned int allmulti; 2088 bool uc_promisc; 2089 #ifdef CONFIG_LOCKDEP 2090 unsigned char nested_level; 2091 #endif 2092 2093 2094 /* Protocol-specific pointers */ 2095 2096 #if IS_ENABLED(CONFIG_VLAN_8021Q) 2097 struct vlan_info __rcu *vlan_info; 2098 #endif 2099 #if IS_ENABLED(CONFIG_NET_DSA) 2100 struct dsa_port *dsa_ptr; 2101 #endif 2102 #if IS_ENABLED(CONFIG_TIPC) 2103 struct tipc_bearer __rcu *tipc_ptr; 2104 #endif 2105 #if IS_ENABLED(CONFIG_ATALK) 2106 void *atalk_ptr; 2107 #endif 2108 struct in_device __rcu *ip_ptr; 2109 #if IS_ENABLED(CONFIG_DECNET) 2110 struct dn_dev __rcu *dn_ptr; 2111 #endif 2112 struct inet6_dev __rcu *ip6_ptr; 2113 #if IS_ENABLED(CONFIG_AX25) 2114 void *ax25_ptr; 2115 #endif 2116 struct wireless_dev *ieee80211_ptr; 2117 struct wpan_dev *ieee802154_ptr; 2118 #if IS_ENABLED(CONFIG_MPLS_ROUTING) 2119 struct mpls_dev __rcu *mpls_ptr; 2120 #endif 2121 #if IS_ENABLED(CONFIG_MCTP) 2122 struct mctp_dev __rcu *mctp_ptr; 2123 #endif 2124 2125 /* 2126 * Cache lines mostly used on receive path (including eth_type_trans()) 2127 */ 2128 /* Interface address info used in eth_type_trans() */ 2129 const unsigned char *dev_addr; 2130 2131 struct netdev_rx_queue *_rx; 2132 unsigned int num_rx_queues; 2133 unsigned int real_num_rx_queues; 2134 2135 struct bpf_prog __rcu *xdp_prog; 2136 unsigned long gro_flush_timeout; 2137 int napi_defer_hard_irqs; 2138 #define GRO_MAX_SIZE 65536 2139 unsigned int gro_max_size; 2140 rx_handler_func_t __rcu *rx_handler; 2141 void __rcu *rx_handler_data; 2142 2143 #ifdef CONFIG_NET_CLS_ACT 2144 struct mini_Qdisc __rcu *miniq_ingress; 2145 #endif 2146 struct netdev_queue __rcu *ingress_queue; 2147 #ifdef CONFIG_NETFILTER_INGRESS 2148 struct nf_hook_entries __rcu *nf_hooks_ingress; 2149 #endif 2150 2151 unsigned char broadcast[MAX_ADDR_LEN]; 2152 #ifdef CONFIG_RFS_ACCEL 2153 struct cpu_rmap *rx_cpu_rmap; 2154 #endif 2155 struct hlist_node index_hlist; 2156 2157 /* 2158 * Cache lines mostly used on transmit path 2159 */ 2160 struct netdev_queue *_tx ____cacheline_aligned_in_smp; 2161 unsigned int num_tx_queues; 2162 unsigned int real_num_tx_queues; 2163 struct Qdisc __rcu *qdisc; 2164 unsigned int tx_queue_len; 2165 spinlock_t tx_global_lock; 2166 2167 struct xdp_dev_bulk_queue __percpu *xdp_bulkq; 2168 2169 #ifdef CONFIG_XPS 2170 struct xps_dev_maps __rcu *xps_maps[XPS_MAPS_MAX]; 2171 #endif 2172 #ifdef CONFIG_NET_CLS_ACT 2173 struct mini_Qdisc __rcu *miniq_egress; 2174 #endif 2175 #ifdef CONFIG_NETFILTER_EGRESS 2176 struct nf_hook_entries __rcu *nf_hooks_egress; 2177 #endif 2178 2179 #ifdef CONFIG_NET_SCHED 2180 DECLARE_HASHTABLE (qdisc_hash, 4); 2181 #endif 2182 /* These may be needed for future network-power-down code. */ 2183 struct timer_list watchdog_timer; 2184 int watchdog_timeo; 2185 2186 u32 proto_down_reason; 2187 2188 struct list_head todo_list; 2189 2190 #ifdef CONFIG_PCPU_DEV_REFCNT 2191 int __percpu *pcpu_refcnt; 2192 #else 2193 refcount_t dev_refcnt; 2194 #endif 2195 struct ref_tracker_dir refcnt_tracker; 2196 2197 struct list_head link_watch_list; 2198 2199 enum { NETREG_UNINITIALIZED=0, 2200 NETREG_REGISTERED, /* completed register_netdevice */ 2201 NETREG_UNREGISTERING, /* called unregister_netdevice */ 2202 NETREG_UNREGISTERED, /* completed unregister todo */ 2203 NETREG_RELEASED, /* called free_netdev */ 2204 NETREG_DUMMY, /* dummy device for NAPI poll */ 2205 } reg_state:8; 2206 2207 bool dismantle; 2208 2209 enum { 2210 RTNL_LINK_INITIALIZED, 2211 RTNL_LINK_INITIALIZING, 2212 } rtnl_link_state:16; 2213 2214 bool needs_free_netdev; 2215 void (*priv_destructor)(struct net_device *dev); 2216 2217 #ifdef CONFIG_NETPOLL 2218 struct netpoll_info __rcu *npinfo; 2219 #endif 2220 2221 possible_net_t nd_net; 2222 2223 /* mid-layer private */ 2224 void *ml_priv; 2225 enum netdev_ml_priv_type ml_priv_type; 2226 2227 union { 2228 struct pcpu_lstats __percpu *lstats; 2229 struct pcpu_sw_netstats __percpu *tstats; 2230 struct pcpu_dstats __percpu *dstats; 2231 }; 2232 2233 #if IS_ENABLED(CONFIG_GARP) 2234 struct garp_port __rcu *garp_port; 2235 #endif 2236 #if IS_ENABLED(CONFIG_MRP) 2237 struct mrp_port __rcu *mrp_port; 2238 #endif 2239 #if IS_ENABLED(CONFIG_NET_DROP_MONITOR) 2240 struct dm_hw_stat_delta __rcu *dm_private; 2241 #endif 2242 struct device dev; 2243 const struct attribute_group *sysfs_groups[4]; 2244 const struct attribute_group *sysfs_rx_queue_group; 2245 2246 const struct rtnl_link_ops *rtnl_link_ops; 2247 2248 /* for setting kernel sock attribute on TCP connection setup */ 2249 #define GSO_MAX_SIZE 65536 2250 unsigned int gso_max_size; 2251 #define GSO_MAX_SEGS 65535 2252 u16 gso_max_segs; 2253 2254 #ifdef CONFIG_DCB 2255 const struct dcbnl_rtnl_ops *dcbnl_ops; 2256 #endif 2257 s16 num_tc; 2258 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE]; 2259 u8 prio_tc_map[TC_BITMASK + 1]; 2260 2261 #if IS_ENABLED(CONFIG_FCOE) 2262 unsigned int fcoe_ddp_xid; 2263 #endif 2264 #if IS_ENABLED(CONFIG_CGROUP_NET_PRIO) 2265 struct netprio_map __rcu *priomap; 2266 #endif 2267 struct phy_device *phydev; 2268 struct sfp_bus *sfp_bus; 2269 struct lock_class_key *qdisc_tx_busylock; 2270 bool proto_down; 2271 unsigned wol_enabled:1; 2272 unsigned threaded:1; 2273 2274 struct list_head net_notifier_list; 2275 2276 #if IS_ENABLED(CONFIG_MACSEC) 2277 /* MACsec management functions */ 2278 const struct macsec_ops *macsec_ops; 2279 #endif 2280 const struct udp_tunnel_nic_info *udp_tunnel_nic_info; 2281 struct udp_tunnel_nic *udp_tunnel_nic; 2282 2283 /* protected by rtnl_lock */ 2284 struct bpf_xdp_entity xdp_state[__MAX_XDP_MODE]; 2285 2286 u8 dev_addr_shadow[MAX_ADDR_LEN]; 2287 netdevice_tracker linkwatch_dev_tracker; 2288 netdevice_tracker watchdog_dev_tracker; 2289 netdevice_tracker dev_registered_tracker; 2290 }; 2291 #define to_net_dev(d) container_of(d, struct net_device, dev) 2292 2293 static inline bool netif_elide_gro(const struct net_device *dev) 2294 { 2295 if (!(dev->features & NETIF_F_GRO) || dev->xdp_prog) 2296 return true; 2297 return false; 2298 } 2299 2300 #define NETDEV_ALIGN 32 2301 2302 static inline 2303 int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio) 2304 { 2305 return dev->prio_tc_map[prio & TC_BITMASK]; 2306 } 2307 2308 static inline 2309 int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc) 2310 { 2311 if (tc >= dev->num_tc) 2312 return -EINVAL; 2313 2314 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK; 2315 return 0; 2316 } 2317 2318 int netdev_txq_to_tc(struct net_device *dev, unsigned int txq); 2319 void netdev_reset_tc(struct net_device *dev); 2320 int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset); 2321 int netdev_set_num_tc(struct net_device *dev, u8 num_tc); 2322 2323 static inline 2324 int netdev_get_num_tc(struct net_device *dev) 2325 { 2326 return dev->num_tc; 2327 } 2328 2329 static inline void net_prefetch(void *p) 2330 { 2331 prefetch(p); 2332 #if L1_CACHE_BYTES < 128 2333 prefetch((u8 *)p + L1_CACHE_BYTES); 2334 #endif 2335 } 2336 2337 static inline void net_prefetchw(void *p) 2338 { 2339 prefetchw(p); 2340 #if L1_CACHE_BYTES < 128 2341 prefetchw((u8 *)p + L1_CACHE_BYTES); 2342 #endif 2343 } 2344 2345 void netdev_unbind_sb_channel(struct net_device *dev, 2346 struct net_device *sb_dev); 2347 int netdev_bind_sb_channel_queue(struct net_device *dev, 2348 struct net_device *sb_dev, 2349 u8 tc, u16 count, u16 offset); 2350 int netdev_set_sb_channel(struct net_device *dev, u16 channel); 2351 static inline int netdev_get_sb_channel(struct net_device *dev) 2352 { 2353 return max_t(int, -dev->num_tc, 0); 2354 } 2355 2356 static inline 2357 struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev, 2358 unsigned int index) 2359 { 2360 return &dev->_tx[index]; 2361 } 2362 2363 static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev, 2364 const struct sk_buff *skb) 2365 { 2366 return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb)); 2367 } 2368 2369 static inline void netdev_for_each_tx_queue(struct net_device *dev, 2370 void (*f)(struct net_device *, 2371 struct netdev_queue *, 2372 void *), 2373 void *arg) 2374 { 2375 unsigned int i; 2376 2377 for (i = 0; i < dev->num_tx_queues; i++) 2378 f(dev, &dev->_tx[i], arg); 2379 } 2380 2381 #define netdev_lockdep_set_classes(dev) \ 2382 { \ 2383 static struct lock_class_key qdisc_tx_busylock_key; \ 2384 static struct lock_class_key qdisc_xmit_lock_key; \ 2385 static struct lock_class_key dev_addr_list_lock_key; \ 2386 unsigned int i; \ 2387 \ 2388 (dev)->qdisc_tx_busylock = &qdisc_tx_busylock_key; \ 2389 lockdep_set_class(&(dev)->addr_list_lock, \ 2390 &dev_addr_list_lock_key); \ 2391 for (i = 0; i < (dev)->num_tx_queues; i++) \ 2392 lockdep_set_class(&(dev)->_tx[i]._xmit_lock, \ 2393 &qdisc_xmit_lock_key); \ 2394 } 2395 2396 u16 netdev_pick_tx(struct net_device *dev, struct sk_buff *skb, 2397 struct net_device *sb_dev); 2398 struct netdev_queue *netdev_core_pick_tx(struct net_device *dev, 2399 struct sk_buff *skb, 2400 struct net_device *sb_dev); 2401 2402 /* returns the headroom that the master device needs to take in account 2403 * when forwarding to this dev 2404 */ 2405 static inline unsigned netdev_get_fwd_headroom(struct net_device *dev) 2406 { 2407 return dev->priv_flags & IFF_PHONY_HEADROOM ? 0 : dev->needed_headroom; 2408 } 2409 2410 static inline void netdev_set_rx_headroom(struct net_device *dev, int new_hr) 2411 { 2412 if (dev->netdev_ops->ndo_set_rx_headroom) 2413 dev->netdev_ops->ndo_set_rx_headroom(dev, new_hr); 2414 } 2415 2416 /* set the device rx headroom to the dev's default */ 2417 static inline void netdev_reset_rx_headroom(struct net_device *dev) 2418 { 2419 netdev_set_rx_headroom(dev, -1); 2420 } 2421 2422 static inline void *netdev_get_ml_priv(struct net_device *dev, 2423 enum netdev_ml_priv_type type) 2424 { 2425 if (dev->ml_priv_type != type) 2426 return NULL; 2427 2428 return dev->ml_priv; 2429 } 2430 2431 static inline void netdev_set_ml_priv(struct net_device *dev, 2432 void *ml_priv, 2433 enum netdev_ml_priv_type type) 2434 { 2435 WARN(dev->ml_priv_type && dev->ml_priv_type != type, 2436 "Overwriting already set ml_priv_type (%u) with different ml_priv_type (%u)!\n", 2437 dev->ml_priv_type, type); 2438 WARN(!dev->ml_priv_type && dev->ml_priv, 2439 "Overwriting already set ml_priv and ml_priv_type is ML_PRIV_NONE!\n"); 2440 2441 dev->ml_priv = ml_priv; 2442 dev->ml_priv_type = type; 2443 } 2444 2445 /* 2446 * Net namespace inlines 2447 */ 2448 static inline 2449 struct net *dev_net(const struct net_device *dev) 2450 { 2451 return read_pnet(&dev->nd_net); 2452 } 2453 2454 static inline 2455 void dev_net_set(struct net_device *dev, struct net *net) 2456 { 2457 write_pnet(&dev->nd_net, net); 2458 } 2459 2460 /** 2461 * netdev_priv - access network device private data 2462 * @dev: network device 2463 * 2464 * Get network device private data 2465 */ 2466 static inline void *netdev_priv(const struct net_device *dev) 2467 { 2468 return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN); 2469 } 2470 2471 /* Set the sysfs physical device reference for the network logical device 2472 * if set prior to registration will cause a symlink during initialization. 2473 */ 2474 #define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev)) 2475 2476 /* Set the sysfs device type for the network logical device to allow 2477 * fine-grained identification of different network device types. For 2478 * example Ethernet, Wireless LAN, Bluetooth, WiMAX etc. 2479 */ 2480 #define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype)) 2481 2482 /* Default NAPI poll() weight 2483 * Device drivers are strongly advised to not use bigger value 2484 */ 2485 #define NAPI_POLL_WEIGHT 64 2486 2487 /** 2488 * netif_napi_add - initialize a NAPI context 2489 * @dev: network device 2490 * @napi: NAPI context 2491 * @poll: polling function 2492 * @weight: default weight 2493 * 2494 * netif_napi_add() must be used to initialize a NAPI context prior to calling 2495 * *any* of the other NAPI-related functions. 2496 */ 2497 void netif_napi_add(struct net_device *dev, struct napi_struct *napi, 2498 int (*poll)(struct napi_struct *, int), int weight); 2499 2500 /** 2501 * netif_tx_napi_add - initialize a NAPI context 2502 * @dev: network device 2503 * @napi: NAPI context 2504 * @poll: polling function 2505 * @weight: default weight 2506 * 2507 * This variant of netif_napi_add() should be used from drivers using NAPI 2508 * to exclusively poll a TX queue. 2509 * This will avoid we add it into napi_hash[], thus polluting this hash table. 2510 */ 2511 static inline void netif_tx_napi_add(struct net_device *dev, 2512 struct napi_struct *napi, 2513 int (*poll)(struct napi_struct *, int), 2514 int weight) 2515 { 2516 set_bit(NAPI_STATE_NO_BUSY_POLL, &napi->state); 2517 netif_napi_add(dev, napi, poll, weight); 2518 } 2519 2520 /** 2521 * __netif_napi_del - remove a NAPI context 2522 * @napi: NAPI context 2523 * 2524 * Warning: caller must observe RCU grace period before freeing memory 2525 * containing @napi. Drivers might want to call this helper to combine 2526 * all the needed RCU grace periods into a single one. 2527 */ 2528 void __netif_napi_del(struct napi_struct *napi); 2529 2530 /** 2531 * netif_napi_del - remove a NAPI context 2532 * @napi: NAPI context 2533 * 2534 * netif_napi_del() removes a NAPI context from the network device NAPI list 2535 */ 2536 static inline void netif_napi_del(struct napi_struct *napi) 2537 { 2538 __netif_napi_del(napi); 2539 synchronize_net(); 2540 } 2541 2542 struct packet_type { 2543 __be16 type; /* This is really htons(ether_type). */ 2544 bool ignore_outgoing; 2545 struct net_device *dev; /* NULL is wildcarded here */ 2546 netdevice_tracker dev_tracker; 2547 int (*func) (struct sk_buff *, 2548 struct net_device *, 2549 struct packet_type *, 2550 struct net_device *); 2551 void (*list_func) (struct list_head *, 2552 struct packet_type *, 2553 struct net_device *); 2554 bool (*id_match)(struct packet_type *ptype, 2555 struct sock *sk); 2556 struct net *af_packet_net; 2557 void *af_packet_priv; 2558 struct list_head list; 2559 }; 2560 2561 struct offload_callbacks { 2562 struct sk_buff *(*gso_segment)(struct sk_buff *skb, 2563 netdev_features_t features); 2564 struct sk_buff *(*gro_receive)(struct list_head *head, 2565 struct sk_buff *skb); 2566 int (*gro_complete)(struct sk_buff *skb, int nhoff); 2567 }; 2568 2569 struct packet_offload { 2570 __be16 type; /* This is really htons(ether_type). */ 2571 u16 priority; 2572 struct offload_callbacks callbacks; 2573 struct list_head list; 2574 }; 2575 2576 /* often modified stats are per-CPU, other are shared (netdev->stats) */ 2577 struct pcpu_sw_netstats { 2578 u64 rx_packets; 2579 u64 rx_bytes; 2580 u64 tx_packets; 2581 u64 tx_bytes; 2582 struct u64_stats_sync syncp; 2583 } __aligned(4 * sizeof(u64)); 2584 2585 struct pcpu_lstats { 2586 u64_stats_t packets; 2587 u64_stats_t bytes; 2588 struct u64_stats_sync syncp; 2589 } __aligned(2 * sizeof(u64)); 2590 2591 void dev_lstats_read(struct net_device *dev, u64 *packets, u64 *bytes); 2592 2593 static inline void dev_sw_netstats_rx_add(struct net_device *dev, 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->rx_bytes += len; 2599 tstats->rx_packets++; 2600 u64_stats_update_end(&tstats->syncp); 2601 } 2602 2603 static inline void dev_sw_netstats_tx_add(struct net_device *dev, 2604 unsigned int packets, 2605 unsigned int len) 2606 { 2607 struct pcpu_sw_netstats *tstats = this_cpu_ptr(dev->tstats); 2608 2609 u64_stats_update_begin(&tstats->syncp); 2610 tstats->tx_bytes += len; 2611 tstats->tx_packets += packets; 2612 u64_stats_update_end(&tstats->syncp); 2613 } 2614 2615 static inline void dev_lstats_add(struct net_device *dev, unsigned int len) 2616 { 2617 struct pcpu_lstats *lstats = this_cpu_ptr(dev->lstats); 2618 2619 u64_stats_update_begin(&lstats->syncp); 2620 u64_stats_add(&lstats->bytes, len); 2621 u64_stats_inc(&lstats->packets); 2622 u64_stats_update_end(&lstats->syncp); 2623 } 2624 2625 #define __netdev_alloc_pcpu_stats(type, gfp) \ 2626 ({ \ 2627 typeof(type) __percpu *pcpu_stats = alloc_percpu_gfp(type, gfp);\ 2628 if (pcpu_stats) { \ 2629 int __cpu; \ 2630 for_each_possible_cpu(__cpu) { \ 2631 typeof(type) *stat; \ 2632 stat = per_cpu_ptr(pcpu_stats, __cpu); \ 2633 u64_stats_init(&stat->syncp); \ 2634 } \ 2635 } \ 2636 pcpu_stats; \ 2637 }) 2638 2639 #define netdev_alloc_pcpu_stats(type) \ 2640 __netdev_alloc_pcpu_stats(type, GFP_KERNEL) 2641 2642 #define devm_netdev_alloc_pcpu_stats(dev, type) \ 2643 ({ \ 2644 typeof(type) __percpu *pcpu_stats = devm_alloc_percpu(dev, type);\ 2645 if (pcpu_stats) { \ 2646 int __cpu; \ 2647 for_each_possible_cpu(__cpu) { \ 2648 typeof(type) *stat; \ 2649 stat = per_cpu_ptr(pcpu_stats, __cpu); \ 2650 u64_stats_init(&stat->syncp); \ 2651 } \ 2652 } \ 2653 pcpu_stats; \ 2654 }) 2655 2656 enum netdev_lag_tx_type { 2657 NETDEV_LAG_TX_TYPE_UNKNOWN, 2658 NETDEV_LAG_TX_TYPE_RANDOM, 2659 NETDEV_LAG_TX_TYPE_BROADCAST, 2660 NETDEV_LAG_TX_TYPE_ROUNDROBIN, 2661 NETDEV_LAG_TX_TYPE_ACTIVEBACKUP, 2662 NETDEV_LAG_TX_TYPE_HASH, 2663 }; 2664 2665 enum netdev_lag_hash { 2666 NETDEV_LAG_HASH_NONE, 2667 NETDEV_LAG_HASH_L2, 2668 NETDEV_LAG_HASH_L34, 2669 NETDEV_LAG_HASH_L23, 2670 NETDEV_LAG_HASH_E23, 2671 NETDEV_LAG_HASH_E34, 2672 NETDEV_LAG_HASH_VLAN_SRCMAC, 2673 NETDEV_LAG_HASH_UNKNOWN, 2674 }; 2675 2676 struct netdev_lag_upper_info { 2677 enum netdev_lag_tx_type tx_type; 2678 enum netdev_lag_hash hash_type; 2679 }; 2680 2681 struct netdev_lag_lower_state_info { 2682 u8 link_up : 1, 2683 tx_enabled : 1; 2684 }; 2685 2686 #include <linux/notifier.h> 2687 2688 /* netdevice notifier chain. Please remember to update netdev_cmd_to_name() 2689 * and the rtnetlink notification exclusion list in rtnetlink_event() when 2690 * adding new types. 2691 */ 2692 enum netdev_cmd { 2693 NETDEV_UP = 1, /* For now you can't veto a device up/down */ 2694 NETDEV_DOWN, 2695 NETDEV_REBOOT, /* Tell a protocol stack a network interface 2696 detected a hardware crash and restarted 2697 - we can use this eg to kick tcp sessions 2698 once done */ 2699 NETDEV_CHANGE, /* Notify device state change */ 2700 NETDEV_REGISTER, 2701 NETDEV_UNREGISTER, 2702 NETDEV_CHANGEMTU, /* notify after mtu change happened */ 2703 NETDEV_CHANGEADDR, /* notify after the address change */ 2704 NETDEV_PRE_CHANGEADDR, /* notify before the address change */ 2705 NETDEV_GOING_DOWN, 2706 NETDEV_CHANGENAME, 2707 NETDEV_FEAT_CHANGE, 2708 NETDEV_BONDING_FAILOVER, 2709 NETDEV_PRE_UP, 2710 NETDEV_PRE_TYPE_CHANGE, 2711 NETDEV_POST_TYPE_CHANGE, 2712 NETDEV_POST_INIT, 2713 NETDEV_RELEASE, 2714 NETDEV_NOTIFY_PEERS, 2715 NETDEV_JOIN, 2716 NETDEV_CHANGEUPPER, 2717 NETDEV_RESEND_IGMP, 2718 NETDEV_PRECHANGEMTU, /* notify before mtu change happened */ 2719 NETDEV_CHANGEINFODATA, 2720 NETDEV_BONDING_INFO, 2721 NETDEV_PRECHANGEUPPER, 2722 NETDEV_CHANGELOWERSTATE, 2723 NETDEV_UDP_TUNNEL_PUSH_INFO, 2724 NETDEV_UDP_TUNNEL_DROP_INFO, 2725 NETDEV_CHANGE_TX_QUEUE_LEN, 2726 NETDEV_CVLAN_FILTER_PUSH_INFO, 2727 NETDEV_CVLAN_FILTER_DROP_INFO, 2728 NETDEV_SVLAN_FILTER_PUSH_INFO, 2729 NETDEV_SVLAN_FILTER_DROP_INFO, 2730 }; 2731 const char *netdev_cmd_to_name(enum netdev_cmd cmd); 2732 2733 int register_netdevice_notifier(struct notifier_block *nb); 2734 int unregister_netdevice_notifier(struct notifier_block *nb); 2735 int register_netdevice_notifier_net(struct net *net, struct notifier_block *nb); 2736 int unregister_netdevice_notifier_net(struct net *net, 2737 struct notifier_block *nb); 2738 int register_netdevice_notifier_dev_net(struct net_device *dev, 2739 struct notifier_block *nb, 2740 struct netdev_net_notifier *nn); 2741 int unregister_netdevice_notifier_dev_net(struct net_device *dev, 2742 struct notifier_block *nb, 2743 struct netdev_net_notifier *nn); 2744 2745 struct netdev_notifier_info { 2746 struct net_device *dev; 2747 struct netlink_ext_ack *extack; 2748 }; 2749 2750 struct netdev_notifier_info_ext { 2751 struct netdev_notifier_info info; /* must be first */ 2752 union { 2753 u32 mtu; 2754 } ext; 2755 }; 2756 2757 struct netdev_notifier_change_info { 2758 struct netdev_notifier_info info; /* must be first */ 2759 unsigned int flags_changed; 2760 }; 2761 2762 struct netdev_notifier_changeupper_info { 2763 struct netdev_notifier_info info; /* must be first */ 2764 struct net_device *upper_dev; /* new upper dev */ 2765 bool master; /* is upper dev master */ 2766 bool linking; /* is the notification for link or unlink */ 2767 void *upper_info; /* upper dev info */ 2768 }; 2769 2770 struct netdev_notifier_changelowerstate_info { 2771 struct netdev_notifier_info info; /* must be first */ 2772 void *lower_state_info; /* is lower dev state */ 2773 }; 2774 2775 struct netdev_notifier_pre_changeaddr_info { 2776 struct netdev_notifier_info info; /* must be first */ 2777 const unsigned char *dev_addr; 2778 }; 2779 2780 static inline void netdev_notifier_info_init(struct netdev_notifier_info *info, 2781 struct net_device *dev) 2782 { 2783 info->dev = dev; 2784 info->extack = NULL; 2785 } 2786 2787 static inline struct net_device * 2788 netdev_notifier_info_to_dev(const struct netdev_notifier_info *info) 2789 { 2790 return info->dev; 2791 } 2792 2793 static inline struct netlink_ext_ack * 2794 netdev_notifier_info_to_extack(const struct netdev_notifier_info *info) 2795 { 2796 return info->extack; 2797 } 2798 2799 int call_netdevice_notifiers(unsigned long val, struct net_device *dev); 2800 2801 2802 extern rwlock_t dev_base_lock; /* Device list lock */ 2803 2804 #define for_each_netdev(net, d) \ 2805 list_for_each_entry(d, &(net)->dev_base_head, dev_list) 2806 #define for_each_netdev_reverse(net, d) \ 2807 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list) 2808 #define for_each_netdev_rcu(net, d) \ 2809 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list) 2810 #define for_each_netdev_safe(net, d, n) \ 2811 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list) 2812 #define for_each_netdev_continue(net, d) \ 2813 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list) 2814 #define for_each_netdev_continue_reverse(net, d) \ 2815 list_for_each_entry_continue_reverse(d, &(net)->dev_base_head, \ 2816 dev_list) 2817 #define for_each_netdev_continue_rcu(net, d) \ 2818 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list) 2819 #define for_each_netdev_in_bond_rcu(bond, slave) \ 2820 for_each_netdev_rcu(&init_net, slave) \ 2821 if (netdev_master_upper_dev_get_rcu(slave) == (bond)) 2822 #define net_device_entry(lh) list_entry(lh, struct net_device, dev_list) 2823 2824 static inline struct net_device *next_net_device(struct net_device *dev) 2825 { 2826 struct list_head *lh; 2827 struct net *net; 2828 2829 net = dev_net(dev); 2830 lh = dev->dev_list.next; 2831 return lh == &net->dev_base_head ? NULL : net_device_entry(lh); 2832 } 2833 2834 static inline struct net_device *next_net_device_rcu(struct net_device *dev) 2835 { 2836 struct list_head *lh; 2837 struct net *net; 2838 2839 net = dev_net(dev); 2840 lh = rcu_dereference(list_next_rcu(&dev->dev_list)); 2841 return lh == &net->dev_base_head ? NULL : net_device_entry(lh); 2842 } 2843 2844 static inline struct net_device *first_net_device(struct net *net) 2845 { 2846 return list_empty(&net->dev_base_head) ? NULL : 2847 net_device_entry(net->dev_base_head.next); 2848 } 2849 2850 static inline struct net_device *first_net_device_rcu(struct net *net) 2851 { 2852 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head)); 2853 2854 return lh == &net->dev_base_head ? NULL : net_device_entry(lh); 2855 } 2856 2857 int netdev_boot_setup_check(struct net_device *dev); 2858 struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type, 2859 const char *hwaddr); 2860 struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type); 2861 void dev_add_pack(struct packet_type *pt); 2862 void dev_remove_pack(struct packet_type *pt); 2863 void __dev_remove_pack(struct packet_type *pt); 2864 void dev_add_offload(struct packet_offload *po); 2865 void dev_remove_offload(struct packet_offload *po); 2866 2867 int dev_get_iflink(const struct net_device *dev); 2868 int dev_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb); 2869 int dev_fill_forward_path(const struct net_device *dev, const u8 *daddr, 2870 struct net_device_path_stack *stack); 2871 struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags, 2872 unsigned short mask); 2873 struct net_device *dev_get_by_name(struct net *net, const char *name); 2874 struct net_device *dev_get_by_name_rcu(struct net *net, const char *name); 2875 struct net_device *__dev_get_by_name(struct net *net, const char *name); 2876 bool netdev_name_in_use(struct net *net, const char *name); 2877 int dev_alloc_name(struct net_device *dev, const char *name); 2878 int dev_open(struct net_device *dev, struct netlink_ext_ack *extack); 2879 void dev_close(struct net_device *dev); 2880 void dev_close_many(struct list_head *head, bool unlink); 2881 void dev_disable_lro(struct net_device *dev); 2882 int dev_loopback_xmit(struct net *net, struct sock *sk, struct sk_buff *newskb); 2883 u16 dev_pick_tx_zero(struct net_device *dev, struct sk_buff *skb, 2884 struct net_device *sb_dev); 2885 u16 dev_pick_tx_cpu_id(struct net_device *dev, struct sk_buff *skb, 2886 struct net_device *sb_dev); 2887 2888 int dev_queue_xmit(struct sk_buff *skb); 2889 int dev_queue_xmit_accel(struct sk_buff *skb, struct net_device *sb_dev); 2890 int __dev_direct_xmit(struct sk_buff *skb, u16 queue_id); 2891 2892 static inline int dev_direct_xmit(struct sk_buff *skb, u16 queue_id) 2893 { 2894 int ret; 2895 2896 ret = __dev_direct_xmit(skb, queue_id); 2897 if (!dev_xmit_complete(ret)) 2898 kfree_skb(skb); 2899 return ret; 2900 } 2901 2902 int register_netdevice(struct net_device *dev); 2903 void unregister_netdevice_queue(struct net_device *dev, struct list_head *head); 2904 void unregister_netdevice_many(struct list_head *head); 2905 static inline void unregister_netdevice(struct net_device *dev) 2906 { 2907 unregister_netdevice_queue(dev, NULL); 2908 } 2909 2910 int netdev_refcnt_read(const struct net_device *dev); 2911 void free_netdev(struct net_device *dev); 2912 void netdev_freemem(struct net_device *dev); 2913 int init_dummy_netdev(struct net_device *dev); 2914 2915 struct net_device *netdev_get_xmit_slave(struct net_device *dev, 2916 struct sk_buff *skb, 2917 bool all_slaves); 2918 struct net_device *netdev_sk_get_lowest_dev(struct net_device *dev, 2919 struct sock *sk); 2920 struct net_device *dev_get_by_index(struct net *net, int ifindex); 2921 struct net_device *__dev_get_by_index(struct net *net, int ifindex); 2922 struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex); 2923 struct net_device *dev_get_by_napi_id(unsigned int napi_id); 2924 int netdev_get_name(struct net *net, char *name, int ifindex); 2925 int dev_restart(struct net_device *dev); 2926 2927 2928 static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev, 2929 unsigned short type, 2930 const void *daddr, const void *saddr, 2931 unsigned int len) 2932 { 2933 if (!dev->header_ops || !dev->header_ops->create) 2934 return 0; 2935 2936 return dev->header_ops->create(skb, dev, type, daddr, saddr, len); 2937 } 2938 2939 static inline int dev_parse_header(const struct sk_buff *skb, 2940 unsigned char *haddr) 2941 { 2942 const struct net_device *dev = skb->dev; 2943 2944 if (!dev->header_ops || !dev->header_ops->parse) 2945 return 0; 2946 return dev->header_ops->parse(skb, haddr); 2947 } 2948 2949 static inline __be16 dev_parse_header_protocol(const struct sk_buff *skb) 2950 { 2951 const struct net_device *dev = skb->dev; 2952 2953 if (!dev->header_ops || !dev->header_ops->parse_protocol) 2954 return 0; 2955 return dev->header_ops->parse_protocol(skb); 2956 } 2957 2958 /* ll_header must have at least hard_header_len allocated */ 2959 static inline bool dev_validate_header(const struct net_device *dev, 2960 char *ll_header, int len) 2961 { 2962 if (likely(len >= dev->hard_header_len)) 2963 return true; 2964 if (len < dev->min_header_len) 2965 return false; 2966 2967 if (capable(CAP_SYS_RAWIO)) { 2968 memset(ll_header + len, 0, dev->hard_header_len - len); 2969 return true; 2970 } 2971 2972 if (dev->header_ops && dev->header_ops->validate) 2973 return dev->header_ops->validate(ll_header, len); 2974 2975 return false; 2976 } 2977 2978 static inline bool dev_has_header(const struct net_device *dev) 2979 { 2980 return dev->header_ops && dev->header_ops->create; 2981 } 2982 2983 #ifdef CONFIG_NET_FLOW_LIMIT 2984 #define FLOW_LIMIT_HISTORY (1 << 7) /* must be ^2 and !overflow buckets */ 2985 struct sd_flow_limit { 2986 u64 count; 2987 unsigned int num_buckets; 2988 unsigned int history_head; 2989 u16 history[FLOW_LIMIT_HISTORY]; 2990 u8 buckets[]; 2991 }; 2992 2993 extern int netdev_flow_limit_table_len; 2994 #endif /* CONFIG_NET_FLOW_LIMIT */ 2995 2996 /* 2997 * Incoming packets are placed on per-CPU queues 2998 */ 2999 struct softnet_data { 3000 struct list_head poll_list; 3001 struct sk_buff_head process_queue; 3002 3003 /* stats */ 3004 unsigned int processed; 3005 unsigned int time_squeeze; 3006 unsigned int received_rps; 3007 #ifdef CONFIG_RPS 3008 struct softnet_data *rps_ipi_list; 3009 #endif 3010 #ifdef CONFIG_NET_FLOW_LIMIT 3011 struct sd_flow_limit __rcu *flow_limit; 3012 #endif 3013 struct Qdisc *output_queue; 3014 struct Qdisc **output_queue_tailp; 3015 struct sk_buff *completion_queue; 3016 #ifdef CONFIG_XFRM_OFFLOAD 3017 struct sk_buff_head xfrm_backlog; 3018 #endif 3019 /* written and read only by owning cpu: */ 3020 struct { 3021 u16 recursion; 3022 u8 more; 3023 } xmit; 3024 #ifdef CONFIG_RPS 3025 /* input_queue_head should be written by cpu owning this struct, 3026 * and only read by other cpus. Worth using a cache line. 3027 */ 3028 unsigned int input_queue_head ____cacheline_aligned_in_smp; 3029 3030 /* Elements below can be accessed between CPUs for RPS/RFS */ 3031 call_single_data_t csd ____cacheline_aligned_in_smp; 3032 struct softnet_data *rps_ipi_next; 3033 unsigned int cpu; 3034 unsigned int input_queue_tail; 3035 #endif 3036 unsigned int dropped; 3037 struct sk_buff_head input_pkt_queue; 3038 struct napi_struct backlog; 3039 3040 }; 3041 3042 static inline void input_queue_head_incr(struct softnet_data *sd) 3043 { 3044 #ifdef CONFIG_RPS 3045 sd->input_queue_head++; 3046 #endif 3047 } 3048 3049 static inline void input_queue_tail_incr_save(struct softnet_data *sd, 3050 unsigned int *qtail) 3051 { 3052 #ifdef CONFIG_RPS 3053 *qtail = ++sd->input_queue_tail; 3054 #endif 3055 } 3056 3057 DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data); 3058 3059 static inline int dev_recursion_level(void) 3060 { 3061 return this_cpu_read(softnet_data.xmit.recursion); 3062 } 3063 3064 #define XMIT_RECURSION_LIMIT 8 3065 static inline bool dev_xmit_recursion(void) 3066 { 3067 return unlikely(__this_cpu_read(softnet_data.xmit.recursion) > 3068 XMIT_RECURSION_LIMIT); 3069 } 3070 3071 static inline void dev_xmit_recursion_inc(void) 3072 { 3073 __this_cpu_inc(softnet_data.xmit.recursion); 3074 } 3075 3076 static inline void dev_xmit_recursion_dec(void) 3077 { 3078 __this_cpu_dec(softnet_data.xmit.recursion); 3079 } 3080 3081 void __netif_schedule(struct Qdisc *q); 3082 void netif_schedule_queue(struct netdev_queue *txq); 3083 3084 static inline void netif_tx_schedule_all(struct net_device *dev) 3085 { 3086 unsigned int i; 3087 3088 for (i = 0; i < dev->num_tx_queues; i++) 3089 netif_schedule_queue(netdev_get_tx_queue(dev, i)); 3090 } 3091 3092 static __always_inline void netif_tx_start_queue(struct netdev_queue *dev_queue) 3093 { 3094 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state); 3095 } 3096 3097 /** 3098 * netif_start_queue - allow transmit 3099 * @dev: network device 3100 * 3101 * Allow upper layers to call the device hard_start_xmit routine. 3102 */ 3103 static inline void netif_start_queue(struct net_device *dev) 3104 { 3105 netif_tx_start_queue(netdev_get_tx_queue(dev, 0)); 3106 } 3107 3108 static inline void netif_tx_start_all_queues(struct net_device *dev) 3109 { 3110 unsigned int i; 3111 3112 for (i = 0; i < dev->num_tx_queues; i++) { 3113 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 3114 netif_tx_start_queue(txq); 3115 } 3116 } 3117 3118 void netif_tx_wake_queue(struct netdev_queue *dev_queue); 3119 3120 /** 3121 * netif_wake_queue - restart transmit 3122 * @dev: network device 3123 * 3124 * Allow upper layers to call the device hard_start_xmit routine. 3125 * Used for flow control when transmit resources are available. 3126 */ 3127 static inline void netif_wake_queue(struct net_device *dev) 3128 { 3129 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0)); 3130 } 3131 3132 static inline void netif_tx_wake_all_queues(struct net_device *dev) 3133 { 3134 unsigned int i; 3135 3136 for (i = 0; i < dev->num_tx_queues; i++) { 3137 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 3138 netif_tx_wake_queue(txq); 3139 } 3140 } 3141 3142 static __always_inline void netif_tx_stop_queue(struct netdev_queue *dev_queue) 3143 { 3144 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state); 3145 } 3146 3147 /** 3148 * netif_stop_queue - stop transmitted packets 3149 * @dev: network device 3150 * 3151 * Stop upper layers calling the device hard_start_xmit routine. 3152 * Used for flow control when transmit resources are unavailable. 3153 */ 3154 static inline void netif_stop_queue(struct net_device *dev) 3155 { 3156 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0)); 3157 } 3158 3159 void netif_tx_stop_all_queues(struct net_device *dev); 3160 3161 static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue) 3162 { 3163 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state); 3164 } 3165 3166 /** 3167 * netif_queue_stopped - test if transmit queue is flowblocked 3168 * @dev: network device 3169 * 3170 * Test if transmit queue on device is currently unable to send. 3171 */ 3172 static inline bool netif_queue_stopped(const struct net_device *dev) 3173 { 3174 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0)); 3175 } 3176 3177 static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue) 3178 { 3179 return dev_queue->state & QUEUE_STATE_ANY_XOFF; 3180 } 3181 3182 static inline bool 3183 netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue) 3184 { 3185 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN; 3186 } 3187 3188 static inline bool 3189 netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue) 3190 { 3191 return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN; 3192 } 3193 3194 /** 3195 * netdev_queue_set_dql_min_limit - set dql minimum limit 3196 * @dev_queue: pointer to transmit queue 3197 * @min_limit: dql minimum limit 3198 * 3199 * Forces xmit_more() to return true until the minimum threshold 3200 * defined by @min_limit is reached (or until the tx queue is 3201 * empty). Warning: to be use with care, misuse will impact the 3202 * latency. 3203 */ 3204 static inline void netdev_queue_set_dql_min_limit(struct netdev_queue *dev_queue, 3205 unsigned int min_limit) 3206 { 3207 #ifdef CONFIG_BQL 3208 dev_queue->dql.min_limit = min_limit; 3209 #endif 3210 } 3211 3212 /** 3213 * netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write 3214 * @dev_queue: pointer to transmit queue 3215 * 3216 * BQL enabled drivers might use this helper in their ndo_start_xmit(), 3217 * to give appropriate hint to the CPU. 3218 */ 3219 static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue) 3220 { 3221 #ifdef CONFIG_BQL 3222 prefetchw(&dev_queue->dql.num_queued); 3223 #endif 3224 } 3225 3226 /** 3227 * netdev_txq_bql_complete_prefetchw - prefetch bql data for write 3228 * @dev_queue: pointer to transmit queue 3229 * 3230 * BQL enabled drivers might use this helper in their TX completion path, 3231 * to give appropriate hint to the CPU. 3232 */ 3233 static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue) 3234 { 3235 #ifdef CONFIG_BQL 3236 prefetchw(&dev_queue->dql.limit); 3237 #endif 3238 } 3239 3240 static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue, 3241 unsigned int bytes) 3242 { 3243 #ifdef CONFIG_BQL 3244 dql_queued(&dev_queue->dql, bytes); 3245 3246 if (likely(dql_avail(&dev_queue->dql) >= 0)) 3247 return; 3248 3249 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state); 3250 3251 /* 3252 * The XOFF flag must be set before checking the dql_avail below, 3253 * because in netdev_tx_completed_queue we update the dql_completed 3254 * before checking the XOFF flag. 3255 */ 3256 smp_mb(); 3257 3258 /* check again in case another CPU has just made room avail */ 3259 if (unlikely(dql_avail(&dev_queue->dql) >= 0)) 3260 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state); 3261 #endif 3262 } 3263 3264 /* Variant of netdev_tx_sent_queue() for drivers that are aware 3265 * that they should not test BQL status themselves. 3266 * We do want to change __QUEUE_STATE_STACK_XOFF only for the last 3267 * skb of a batch. 3268 * Returns true if the doorbell must be used to kick the NIC. 3269 */ 3270 static inline bool __netdev_tx_sent_queue(struct netdev_queue *dev_queue, 3271 unsigned int bytes, 3272 bool xmit_more) 3273 { 3274 if (xmit_more) { 3275 #ifdef CONFIG_BQL 3276 dql_queued(&dev_queue->dql, bytes); 3277 #endif 3278 return netif_tx_queue_stopped(dev_queue); 3279 } 3280 netdev_tx_sent_queue(dev_queue, bytes); 3281 return true; 3282 } 3283 3284 /** 3285 * netdev_sent_queue - report the number of bytes queued to hardware 3286 * @dev: network device 3287 * @bytes: number of bytes queued to the hardware device queue 3288 * 3289 * Report the number of bytes queued for sending/completion to the network 3290 * device hardware queue. @bytes should be a good approximation and should 3291 * exactly match netdev_completed_queue() @bytes 3292 */ 3293 static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes) 3294 { 3295 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes); 3296 } 3297 3298 static inline bool __netdev_sent_queue(struct net_device *dev, 3299 unsigned int bytes, 3300 bool xmit_more) 3301 { 3302 return __netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes, 3303 xmit_more); 3304 } 3305 3306 static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue, 3307 unsigned int pkts, unsigned int bytes) 3308 { 3309 #ifdef CONFIG_BQL 3310 if (unlikely(!bytes)) 3311 return; 3312 3313 dql_completed(&dev_queue->dql, bytes); 3314 3315 /* 3316 * Without the memory barrier there is a small possiblity that 3317 * netdev_tx_sent_queue will miss the update and cause the queue to 3318 * be stopped forever 3319 */ 3320 smp_mb(); 3321 3322 if (unlikely(dql_avail(&dev_queue->dql) < 0)) 3323 return; 3324 3325 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state)) 3326 netif_schedule_queue(dev_queue); 3327 #endif 3328 } 3329 3330 /** 3331 * netdev_completed_queue - report bytes and packets completed by device 3332 * @dev: network device 3333 * @pkts: actual number of packets sent over the medium 3334 * @bytes: actual number of bytes sent over the medium 3335 * 3336 * Report the number of bytes and packets transmitted by the network device 3337 * hardware queue over the physical medium, @bytes must exactly match the 3338 * @bytes amount passed to netdev_sent_queue() 3339 */ 3340 static inline void netdev_completed_queue(struct net_device *dev, 3341 unsigned int pkts, unsigned int bytes) 3342 { 3343 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes); 3344 } 3345 3346 static inline void netdev_tx_reset_queue(struct netdev_queue *q) 3347 { 3348 #ifdef CONFIG_BQL 3349 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state); 3350 dql_reset(&q->dql); 3351 #endif 3352 } 3353 3354 /** 3355 * netdev_reset_queue - reset the packets and bytes count of a network device 3356 * @dev_queue: network device 3357 * 3358 * Reset the bytes and packet count of a network device and clear the 3359 * software flow control OFF bit for this network device 3360 */ 3361 static inline void netdev_reset_queue(struct net_device *dev_queue) 3362 { 3363 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0)); 3364 } 3365 3366 /** 3367 * netdev_cap_txqueue - check if selected tx queue exceeds device queues 3368 * @dev: network device 3369 * @queue_index: given tx queue index 3370 * 3371 * Returns 0 if given tx queue index >= number of device tx queues, 3372 * otherwise returns the originally passed tx queue index. 3373 */ 3374 static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index) 3375 { 3376 if (unlikely(queue_index >= dev->real_num_tx_queues)) { 3377 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n", 3378 dev->name, queue_index, 3379 dev->real_num_tx_queues); 3380 return 0; 3381 } 3382 3383 return queue_index; 3384 } 3385 3386 /** 3387 * netif_running - test if up 3388 * @dev: network device 3389 * 3390 * Test if the device has been brought up. 3391 */ 3392 static inline bool netif_running(const struct net_device *dev) 3393 { 3394 return test_bit(__LINK_STATE_START, &dev->state); 3395 } 3396 3397 /* 3398 * Routines to manage the subqueues on a device. We only need start, 3399 * stop, and a check if it's stopped. All other device management is 3400 * done at the overall netdevice level. 3401 * Also test the device if we're multiqueue. 3402 */ 3403 3404 /** 3405 * netif_start_subqueue - allow sending packets on subqueue 3406 * @dev: network device 3407 * @queue_index: sub queue index 3408 * 3409 * Start individual transmit queue of a device with multiple transmit queues. 3410 */ 3411 static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index) 3412 { 3413 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 3414 3415 netif_tx_start_queue(txq); 3416 } 3417 3418 /** 3419 * netif_stop_subqueue - stop sending packets on subqueue 3420 * @dev: network device 3421 * @queue_index: sub queue index 3422 * 3423 * Stop individual transmit queue of a device with multiple transmit queues. 3424 */ 3425 static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index) 3426 { 3427 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 3428 netif_tx_stop_queue(txq); 3429 } 3430 3431 /** 3432 * __netif_subqueue_stopped - test status of subqueue 3433 * @dev: network device 3434 * @queue_index: sub queue index 3435 * 3436 * Check individual transmit queue of a device with multiple transmit queues. 3437 */ 3438 static inline bool __netif_subqueue_stopped(const struct net_device *dev, 3439 u16 queue_index) 3440 { 3441 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 3442 3443 return netif_tx_queue_stopped(txq); 3444 } 3445 3446 /** 3447 * netif_subqueue_stopped - test status of subqueue 3448 * @dev: network device 3449 * @skb: sub queue buffer pointer 3450 * 3451 * Check individual transmit queue of a device with multiple transmit queues. 3452 */ 3453 static inline bool netif_subqueue_stopped(const struct net_device *dev, 3454 struct sk_buff *skb) 3455 { 3456 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb)); 3457 } 3458 3459 /** 3460 * netif_wake_subqueue - allow sending packets on subqueue 3461 * @dev: network device 3462 * @queue_index: sub queue index 3463 * 3464 * Resume individual transmit queue of a device with multiple transmit queues. 3465 */ 3466 static inline void netif_wake_subqueue(struct net_device *dev, u16 queue_index) 3467 { 3468 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 3469 3470 netif_tx_wake_queue(txq); 3471 } 3472 3473 #ifdef CONFIG_XPS 3474 int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask, 3475 u16 index); 3476 int __netif_set_xps_queue(struct net_device *dev, const unsigned long *mask, 3477 u16 index, enum xps_map_type type); 3478 3479 /** 3480 * netif_attr_test_mask - Test a CPU or Rx queue set in a mask 3481 * @j: CPU/Rx queue index 3482 * @mask: bitmask of all cpus/rx queues 3483 * @nr_bits: number of bits in the bitmask 3484 * 3485 * Test if a CPU or Rx queue index is set in a mask of all CPU/Rx queues. 3486 */ 3487 static inline bool netif_attr_test_mask(unsigned long j, 3488 const unsigned long *mask, 3489 unsigned int nr_bits) 3490 { 3491 cpu_max_bits_warn(j, nr_bits); 3492 return test_bit(j, mask); 3493 } 3494 3495 /** 3496 * netif_attr_test_online - Test for online CPU/Rx queue 3497 * @j: CPU/Rx queue index 3498 * @online_mask: bitmask for CPUs/Rx queues that are online 3499 * @nr_bits: number of bits in the bitmask 3500 * 3501 * Returns true if a CPU/Rx queue is online. 3502 */ 3503 static inline bool netif_attr_test_online(unsigned long j, 3504 const unsigned long *online_mask, 3505 unsigned int nr_bits) 3506 { 3507 cpu_max_bits_warn(j, nr_bits); 3508 3509 if (online_mask) 3510 return test_bit(j, online_mask); 3511 3512 return (j < nr_bits); 3513 } 3514 3515 /** 3516 * netif_attrmask_next - get the next CPU/Rx queue in a cpu/Rx queues mask 3517 * @n: CPU/Rx queue index 3518 * @srcp: the cpumask/Rx queue mask pointer 3519 * @nr_bits: number of bits in the bitmask 3520 * 3521 * Returns >= nr_bits if no further CPUs/Rx queues set. 3522 */ 3523 static inline unsigned int netif_attrmask_next(int n, const unsigned long *srcp, 3524 unsigned int nr_bits) 3525 { 3526 /* -1 is a legal arg here. */ 3527 if (n != -1) 3528 cpu_max_bits_warn(n, nr_bits); 3529 3530 if (srcp) 3531 return find_next_bit(srcp, nr_bits, n + 1); 3532 3533 return n + 1; 3534 } 3535 3536 /** 3537 * netif_attrmask_next_and - get the next CPU/Rx queue in \*src1p & \*src2p 3538 * @n: CPU/Rx queue index 3539 * @src1p: the first CPUs/Rx queues mask pointer 3540 * @src2p: the second CPUs/Rx queues mask pointer 3541 * @nr_bits: number of bits in the bitmask 3542 * 3543 * Returns >= nr_bits if no further CPUs/Rx queues set in both. 3544 */ 3545 static inline int netif_attrmask_next_and(int n, const unsigned long *src1p, 3546 const unsigned long *src2p, 3547 unsigned int nr_bits) 3548 { 3549 /* -1 is a legal arg here. */ 3550 if (n != -1) 3551 cpu_max_bits_warn(n, nr_bits); 3552 3553 if (src1p && src2p) 3554 return find_next_and_bit(src1p, src2p, nr_bits, n + 1); 3555 else if (src1p) 3556 return find_next_bit(src1p, nr_bits, n + 1); 3557 else if (src2p) 3558 return find_next_bit(src2p, nr_bits, n + 1); 3559 3560 return n + 1; 3561 } 3562 #else 3563 static inline int netif_set_xps_queue(struct net_device *dev, 3564 const struct cpumask *mask, 3565 u16 index) 3566 { 3567 return 0; 3568 } 3569 3570 static inline int __netif_set_xps_queue(struct net_device *dev, 3571 const unsigned long *mask, 3572 u16 index, enum xps_map_type type) 3573 { 3574 return 0; 3575 } 3576 #endif 3577 3578 /** 3579 * netif_is_multiqueue - test if device has multiple transmit queues 3580 * @dev: network device 3581 * 3582 * Check if device has multiple transmit queues 3583 */ 3584 static inline bool netif_is_multiqueue(const struct net_device *dev) 3585 { 3586 return dev->num_tx_queues > 1; 3587 } 3588 3589 int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq); 3590 3591 #ifdef CONFIG_SYSFS 3592 int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq); 3593 #else 3594 static inline int netif_set_real_num_rx_queues(struct net_device *dev, 3595 unsigned int rxqs) 3596 { 3597 dev->real_num_rx_queues = rxqs; 3598 return 0; 3599 } 3600 #endif 3601 int netif_set_real_num_queues(struct net_device *dev, 3602 unsigned int txq, unsigned int rxq); 3603 3604 static inline struct netdev_rx_queue * 3605 __netif_get_rx_queue(struct net_device *dev, unsigned int rxq) 3606 { 3607 return dev->_rx + rxq; 3608 } 3609 3610 #ifdef CONFIG_SYSFS 3611 static inline unsigned int get_netdev_rx_queue_index( 3612 struct netdev_rx_queue *queue) 3613 { 3614 struct net_device *dev = queue->dev; 3615 int index = queue - dev->_rx; 3616 3617 BUG_ON(index >= dev->num_rx_queues); 3618 return index; 3619 } 3620 #endif 3621 3622 #define DEFAULT_MAX_NUM_RSS_QUEUES (8) 3623 int netif_get_num_default_rss_queues(void); 3624 3625 enum skb_free_reason { 3626 SKB_REASON_CONSUMED, 3627 SKB_REASON_DROPPED, 3628 }; 3629 3630 void __dev_kfree_skb_irq(struct sk_buff *skb, enum skb_free_reason reason); 3631 void __dev_kfree_skb_any(struct sk_buff *skb, enum skb_free_reason reason); 3632 3633 /* 3634 * It is not allowed to call kfree_skb() or consume_skb() from hardware 3635 * interrupt context or with hardware interrupts being disabled. 3636 * (in_hardirq() || irqs_disabled()) 3637 * 3638 * We provide four helpers that can be used in following contexts : 3639 * 3640 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context, 3641 * replacing kfree_skb(skb) 3642 * 3643 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context. 3644 * Typically used in place of consume_skb(skb) in TX completion path 3645 * 3646 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context, 3647 * replacing kfree_skb(skb) 3648 * 3649 * dev_consume_skb_any(skb) when caller doesn't know its current irq context, 3650 * and consumed a packet. Used in place of consume_skb(skb) 3651 */ 3652 static inline void dev_kfree_skb_irq(struct sk_buff *skb) 3653 { 3654 __dev_kfree_skb_irq(skb, SKB_REASON_DROPPED); 3655 } 3656 3657 static inline void dev_consume_skb_irq(struct sk_buff *skb) 3658 { 3659 __dev_kfree_skb_irq(skb, SKB_REASON_CONSUMED); 3660 } 3661 3662 static inline void dev_kfree_skb_any(struct sk_buff *skb) 3663 { 3664 __dev_kfree_skb_any(skb, SKB_REASON_DROPPED); 3665 } 3666 3667 static inline void dev_consume_skb_any(struct sk_buff *skb) 3668 { 3669 __dev_kfree_skb_any(skb, SKB_REASON_CONSUMED); 3670 } 3671 3672 u32 bpf_prog_run_generic_xdp(struct sk_buff *skb, struct xdp_buff *xdp, 3673 struct bpf_prog *xdp_prog); 3674 void generic_xdp_tx(struct sk_buff *skb, struct bpf_prog *xdp_prog); 3675 int do_xdp_generic(struct bpf_prog *xdp_prog, struct sk_buff *skb); 3676 int netif_rx(struct sk_buff *skb); 3677 int __netif_rx(struct sk_buff *skb); 3678 3679 static inline int netif_rx_ni(struct sk_buff *skb) 3680 { 3681 return netif_rx(skb); 3682 } 3683 3684 static inline int netif_rx_any_context(struct sk_buff *skb) 3685 { 3686 return netif_rx(skb); 3687 } 3688 3689 int netif_receive_skb(struct sk_buff *skb); 3690 int netif_receive_skb_core(struct sk_buff *skb); 3691 void netif_receive_skb_list_internal(struct list_head *head); 3692 void netif_receive_skb_list(struct list_head *head); 3693 gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb); 3694 void napi_gro_flush(struct napi_struct *napi, bool flush_old); 3695 struct sk_buff *napi_get_frags(struct napi_struct *napi); 3696 gro_result_t napi_gro_frags(struct napi_struct *napi); 3697 struct packet_offload *gro_find_receive_by_type(__be16 type); 3698 struct packet_offload *gro_find_complete_by_type(__be16 type); 3699 3700 static inline void napi_free_frags(struct napi_struct *napi) 3701 { 3702 kfree_skb(napi->skb); 3703 napi->skb = NULL; 3704 } 3705 3706 bool netdev_is_rx_handler_busy(struct net_device *dev); 3707 int netdev_rx_handler_register(struct net_device *dev, 3708 rx_handler_func_t *rx_handler, 3709 void *rx_handler_data); 3710 void netdev_rx_handler_unregister(struct net_device *dev); 3711 3712 bool dev_valid_name(const char *name); 3713 static inline bool is_socket_ioctl_cmd(unsigned int cmd) 3714 { 3715 return _IOC_TYPE(cmd) == SOCK_IOC_TYPE; 3716 } 3717 int get_user_ifreq(struct ifreq *ifr, void __user **ifrdata, void __user *arg); 3718 int put_user_ifreq(struct ifreq *ifr, void __user *arg); 3719 int dev_ioctl(struct net *net, unsigned int cmd, struct ifreq *ifr, 3720 void __user *data, bool *need_copyout); 3721 int dev_ifconf(struct net *net, struct ifconf __user *ifc); 3722 int dev_ethtool(struct net *net, struct ifreq *ifr, void __user *userdata); 3723 unsigned int dev_get_flags(const struct net_device *); 3724 int __dev_change_flags(struct net_device *dev, unsigned int flags, 3725 struct netlink_ext_ack *extack); 3726 int dev_change_flags(struct net_device *dev, unsigned int flags, 3727 struct netlink_ext_ack *extack); 3728 void __dev_notify_flags(struct net_device *, unsigned int old_flags, 3729 unsigned int gchanges); 3730 int dev_change_name(struct net_device *, const char *); 3731 int dev_set_alias(struct net_device *, const char *, size_t); 3732 int dev_get_alias(const struct net_device *, char *, size_t); 3733 int __dev_change_net_namespace(struct net_device *dev, struct net *net, 3734 const char *pat, int new_ifindex); 3735 static inline 3736 int dev_change_net_namespace(struct net_device *dev, struct net *net, 3737 const char *pat) 3738 { 3739 return __dev_change_net_namespace(dev, net, pat, 0); 3740 } 3741 int __dev_set_mtu(struct net_device *, int); 3742 int dev_validate_mtu(struct net_device *dev, int mtu, 3743 struct netlink_ext_ack *extack); 3744 int dev_set_mtu_ext(struct net_device *dev, int mtu, 3745 struct netlink_ext_ack *extack); 3746 int dev_set_mtu(struct net_device *, int); 3747 int dev_change_tx_queue_len(struct net_device *, unsigned long); 3748 void dev_set_group(struct net_device *, int); 3749 int dev_pre_changeaddr_notify(struct net_device *dev, const char *addr, 3750 struct netlink_ext_ack *extack); 3751 int dev_set_mac_address(struct net_device *dev, struct sockaddr *sa, 3752 struct netlink_ext_ack *extack); 3753 int dev_set_mac_address_user(struct net_device *dev, struct sockaddr *sa, 3754 struct netlink_ext_ack *extack); 3755 int dev_get_mac_address(struct sockaddr *sa, struct net *net, char *dev_name); 3756 int dev_change_carrier(struct net_device *, bool new_carrier); 3757 int dev_get_phys_port_id(struct net_device *dev, 3758 struct netdev_phys_item_id *ppid); 3759 int dev_get_phys_port_name(struct net_device *dev, 3760 char *name, size_t len); 3761 int dev_get_port_parent_id(struct net_device *dev, 3762 struct netdev_phys_item_id *ppid, bool recurse); 3763 bool netdev_port_same_parent_id(struct net_device *a, struct net_device *b); 3764 int dev_change_proto_down(struct net_device *dev, bool proto_down); 3765 void dev_change_proto_down_reason(struct net_device *dev, unsigned long mask, 3766 u32 value); 3767 struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev, bool *again); 3768 struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev, 3769 struct netdev_queue *txq, int *ret); 3770 3771 typedef int (*bpf_op_t)(struct net_device *dev, struct netdev_bpf *bpf); 3772 int dev_change_xdp_fd(struct net_device *dev, struct netlink_ext_ack *extack, 3773 int fd, int expected_fd, u32 flags); 3774 int bpf_xdp_link_attach(const union bpf_attr *attr, struct bpf_prog *prog); 3775 u8 dev_xdp_prog_count(struct net_device *dev); 3776 u32 dev_xdp_prog_id(struct net_device *dev, enum bpf_xdp_mode mode); 3777 3778 int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb); 3779 int dev_forward_skb(struct net_device *dev, struct sk_buff *skb); 3780 int dev_forward_skb_nomtu(struct net_device *dev, struct sk_buff *skb); 3781 bool is_skb_forwardable(const struct net_device *dev, 3782 const struct sk_buff *skb); 3783 3784 static __always_inline bool __is_skb_forwardable(const struct net_device *dev, 3785 const struct sk_buff *skb, 3786 const bool check_mtu) 3787 { 3788 const u32 vlan_hdr_len = 4; /* VLAN_HLEN */ 3789 unsigned int len; 3790 3791 if (!(dev->flags & IFF_UP)) 3792 return false; 3793 3794 if (!check_mtu) 3795 return true; 3796 3797 len = dev->mtu + dev->hard_header_len + vlan_hdr_len; 3798 if (skb->len <= len) 3799 return true; 3800 3801 /* if TSO is enabled, we don't care about the length as the packet 3802 * could be forwarded without being segmented before 3803 */ 3804 if (skb_is_gso(skb)) 3805 return true; 3806 3807 return false; 3808 } 3809 3810 static __always_inline int ____dev_forward_skb(struct net_device *dev, 3811 struct sk_buff *skb, 3812 const bool check_mtu) 3813 { 3814 if (skb_orphan_frags(skb, GFP_ATOMIC) || 3815 unlikely(!__is_skb_forwardable(dev, skb, check_mtu))) { 3816 atomic_long_inc(&dev->rx_dropped); 3817 kfree_skb(skb); 3818 return NET_RX_DROP; 3819 } 3820 3821 skb_scrub_packet(skb, !net_eq(dev_net(dev), dev_net(skb->dev))); 3822 skb->priority = 0; 3823 return 0; 3824 } 3825 3826 bool dev_nit_active(struct net_device *dev); 3827 void dev_queue_xmit_nit(struct sk_buff *skb, struct net_device *dev); 3828 3829 extern int netdev_budget; 3830 extern unsigned int netdev_budget_usecs; 3831 3832 /* Called by rtnetlink.c:rtnl_unlock() */ 3833 void netdev_run_todo(void); 3834 3835 static inline void __dev_put(struct net_device *dev) 3836 { 3837 if (dev) { 3838 #ifdef CONFIG_PCPU_DEV_REFCNT 3839 this_cpu_dec(*dev->pcpu_refcnt); 3840 #else 3841 refcount_dec(&dev->dev_refcnt); 3842 #endif 3843 } 3844 } 3845 3846 static inline void __dev_hold(struct net_device *dev) 3847 { 3848 if (dev) { 3849 #ifdef CONFIG_PCPU_DEV_REFCNT 3850 this_cpu_inc(*dev->pcpu_refcnt); 3851 #else 3852 refcount_inc(&dev->dev_refcnt); 3853 #endif 3854 } 3855 } 3856 3857 static inline void __netdev_tracker_alloc(struct net_device *dev, 3858 netdevice_tracker *tracker, 3859 gfp_t gfp) 3860 { 3861 #ifdef CONFIG_NET_DEV_REFCNT_TRACKER 3862 ref_tracker_alloc(&dev->refcnt_tracker, tracker, gfp); 3863 #endif 3864 } 3865 3866 /* netdev_tracker_alloc() can upgrade a prior untracked reference 3867 * taken by dev_get_by_name()/dev_get_by_index() to a tracked one. 3868 */ 3869 static inline void netdev_tracker_alloc(struct net_device *dev, 3870 netdevice_tracker *tracker, gfp_t gfp) 3871 { 3872 #ifdef CONFIG_NET_DEV_REFCNT_TRACKER 3873 refcount_dec(&dev->refcnt_tracker.no_tracker); 3874 __netdev_tracker_alloc(dev, tracker, gfp); 3875 #endif 3876 } 3877 3878 static inline void netdev_tracker_free(struct net_device *dev, 3879 netdevice_tracker *tracker) 3880 { 3881 #ifdef CONFIG_NET_DEV_REFCNT_TRACKER 3882 ref_tracker_free(&dev->refcnt_tracker, tracker); 3883 #endif 3884 } 3885 3886 static inline void dev_hold_track(struct net_device *dev, 3887 netdevice_tracker *tracker, gfp_t gfp) 3888 { 3889 if (dev) { 3890 __dev_hold(dev); 3891 __netdev_tracker_alloc(dev, tracker, gfp); 3892 } 3893 } 3894 3895 static inline void dev_put_track(struct net_device *dev, 3896 netdevice_tracker *tracker) 3897 { 3898 if (dev) { 3899 netdev_tracker_free(dev, tracker); 3900 __dev_put(dev); 3901 } 3902 } 3903 3904 /** 3905 * dev_hold - get reference to device 3906 * @dev: network device 3907 * 3908 * Hold reference to device to keep it from being freed. 3909 * Try using dev_hold_track() instead. 3910 */ 3911 static inline void dev_hold(struct net_device *dev) 3912 { 3913 dev_hold_track(dev, NULL, GFP_ATOMIC); 3914 } 3915 3916 /** 3917 * dev_put - release reference to device 3918 * @dev: network device 3919 * 3920 * Release reference to device to allow it to be freed. 3921 * Try using dev_put_track() instead. 3922 */ 3923 static inline void dev_put(struct net_device *dev) 3924 { 3925 dev_put_track(dev, NULL); 3926 } 3927 3928 static inline void dev_replace_track(struct net_device *odev, 3929 struct net_device *ndev, 3930 netdevice_tracker *tracker, 3931 gfp_t gfp) 3932 { 3933 if (odev) 3934 netdev_tracker_free(odev, tracker); 3935 3936 __dev_hold(ndev); 3937 __dev_put(odev); 3938 3939 if (ndev) 3940 __netdev_tracker_alloc(ndev, tracker, gfp); 3941 } 3942 3943 /* Carrier loss detection, dial on demand. The functions netif_carrier_on 3944 * and _off may be called from IRQ context, but it is caller 3945 * who is responsible for serialization of these calls. 3946 * 3947 * The name carrier is inappropriate, these functions should really be 3948 * called netif_lowerlayer_*() because they represent the state of any 3949 * kind of lower layer not just hardware media. 3950 */ 3951 3952 void linkwatch_init_dev(struct net_device *dev); 3953 void linkwatch_fire_event(struct net_device *dev); 3954 void linkwatch_forget_dev(struct net_device *dev); 3955 3956 /** 3957 * netif_carrier_ok - test if carrier present 3958 * @dev: network device 3959 * 3960 * Check if carrier is present on device 3961 */ 3962 static inline bool netif_carrier_ok(const struct net_device *dev) 3963 { 3964 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state); 3965 } 3966 3967 unsigned long dev_trans_start(struct net_device *dev); 3968 3969 void __netdev_watchdog_up(struct net_device *dev); 3970 3971 void netif_carrier_on(struct net_device *dev); 3972 void netif_carrier_off(struct net_device *dev); 3973 void netif_carrier_event(struct net_device *dev); 3974 3975 /** 3976 * netif_dormant_on - mark device as dormant. 3977 * @dev: network device 3978 * 3979 * Mark device as dormant (as per RFC2863). 3980 * 3981 * The dormant state indicates that the relevant interface is not 3982 * actually in a condition to pass packets (i.e., it is not 'up') but is 3983 * in a "pending" state, waiting for some external event. For "on- 3984 * demand" interfaces, this new state identifies the situation where the 3985 * interface is waiting for events to place it in the up state. 3986 */ 3987 static inline void netif_dormant_on(struct net_device *dev) 3988 { 3989 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state)) 3990 linkwatch_fire_event(dev); 3991 } 3992 3993 /** 3994 * netif_dormant_off - set device as not dormant. 3995 * @dev: network device 3996 * 3997 * Device is not in dormant state. 3998 */ 3999 static inline void netif_dormant_off(struct net_device *dev) 4000 { 4001 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state)) 4002 linkwatch_fire_event(dev); 4003 } 4004 4005 /** 4006 * netif_dormant - test if device is dormant 4007 * @dev: network device 4008 * 4009 * Check if device is dormant. 4010 */ 4011 static inline bool netif_dormant(const struct net_device *dev) 4012 { 4013 return test_bit(__LINK_STATE_DORMANT, &dev->state); 4014 } 4015 4016 4017 /** 4018 * netif_testing_on - mark device as under test. 4019 * @dev: network device 4020 * 4021 * Mark device as under test (as per RFC2863). 4022 * 4023 * The testing state indicates that some test(s) must be performed on 4024 * the interface. After completion, of the test, the interface state 4025 * will change to up, dormant, or down, as appropriate. 4026 */ 4027 static inline void netif_testing_on(struct net_device *dev) 4028 { 4029 if (!test_and_set_bit(__LINK_STATE_TESTING, &dev->state)) 4030 linkwatch_fire_event(dev); 4031 } 4032 4033 /** 4034 * netif_testing_off - set device as not under test. 4035 * @dev: network device 4036 * 4037 * Device is not in testing state. 4038 */ 4039 static inline void netif_testing_off(struct net_device *dev) 4040 { 4041 if (test_and_clear_bit(__LINK_STATE_TESTING, &dev->state)) 4042 linkwatch_fire_event(dev); 4043 } 4044 4045 /** 4046 * netif_testing - test if device is under test 4047 * @dev: network device 4048 * 4049 * Check if device is under test 4050 */ 4051 static inline bool netif_testing(const struct net_device *dev) 4052 { 4053 return test_bit(__LINK_STATE_TESTING, &dev->state); 4054 } 4055 4056 4057 /** 4058 * netif_oper_up - test if device is operational 4059 * @dev: network device 4060 * 4061 * Check if carrier is operational 4062 */ 4063 static inline bool netif_oper_up(const struct net_device *dev) 4064 { 4065 return (dev->operstate == IF_OPER_UP || 4066 dev->operstate == IF_OPER_UNKNOWN /* backward compat */); 4067 } 4068 4069 /** 4070 * netif_device_present - is device available or removed 4071 * @dev: network device 4072 * 4073 * Check if device has not been removed from system. 4074 */ 4075 static inline bool netif_device_present(const struct net_device *dev) 4076 { 4077 return test_bit(__LINK_STATE_PRESENT, &dev->state); 4078 } 4079 4080 void netif_device_detach(struct net_device *dev); 4081 4082 void netif_device_attach(struct net_device *dev); 4083 4084 /* 4085 * Network interface message level settings 4086 */ 4087 4088 enum { 4089 NETIF_MSG_DRV_BIT, 4090 NETIF_MSG_PROBE_BIT, 4091 NETIF_MSG_LINK_BIT, 4092 NETIF_MSG_TIMER_BIT, 4093 NETIF_MSG_IFDOWN_BIT, 4094 NETIF_MSG_IFUP_BIT, 4095 NETIF_MSG_RX_ERR_BIT, 4096 NETIF_MSG_TX_ERR_BIT, 4097 NETIF_MSG_TX_QUEUED_BIT, 4098 NETIF_MSG_INTR_BIT, 4099 NETIF_MSG_TX_DONE_BIT, 4100 NETIF_MSG_RX_STATUS_BIT, 4101 NETIF_MSG_PKTDATA_BIT, 4102 NETIF_MSG_HW_BIT, 4103 NETIF_MSG_WOL_BIT, 4104 4105 /* When you add a new bit above, update netif_msg_class_names array 4106 * in net/ethtool/common.c 4107 */ 4108 NETIF_MSG_CLASS_COUNT, 4109 }; 4110 /* Both ethtool_ops interface and internal driver implementation use u32 */ 4111 static_assert(NETIF_MSG_CLASS_COUNT <= 32); 4112 4113 #define __NETIF_MSG_BIT(bit) ((u32)1 << (bit)) 4114 #define __NETIF_MSG(name) __NETIF_MSG_BIT(NETIF_MSG_ ## name ## _BIT) 4115 4116 #define NETIF_MSG_DRV __NETIF_MSG(DRV) 4117 #define NETIF_MSG_PROBE __NETIF_MSG(PROBE) 4118 #define NETIF_MSG_LINK __NETIF_MSG(LINK) 4119 #define NETIF_MSG_TIMER __NETIF_MSG(TIMER) 4120 #define NETIF_MSG_IFDOWN __NETIF_MSG(IFDOWN) 4121 #define NETIF_MSG_IFUP __NETIF_MSG(IFUP) 4122 #define NETIF_MSG_RX_ERR __NETIF_MSG(RX_ERR) 4123 #define NETIF_MSG_TX_ERR __NETIF_MSG(TX_ERR) 4124 #define NETIF_MSG_TX_QUEUED __NETIF_MSG(TX_QUEUED) 4125 #define NETIF_MSG_INTR __NETIF_MSG(INTR) 4126 #define NETIF_MSG_TX_DONE __NETIF_MSG(TX_DONE) 4127 #define NETIF_MSG_RX_STATUS __NETIF_MSG(RX_STATUS) 4128 #define NETIF_MSG_PKTDATA __NETIF_MSG(PKTDATA) 4129 #define NETIF_MSG_HW __NETIF_MSG(HW) 4130 #define NETIF_MSG_WOL __NETIF_MSG(WOL) 4131 4132 #define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV) 4133 #define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE) 4134 #define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK) 4135 #define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER) 4136 #define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN) 4137 #define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP) 4138 #define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR) 4139 #define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR) 4140 #define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED) 4141 #define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR) 4142 #define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE) 4143 #define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS) 4144 #define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA) 4145 #define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW) 4146 #define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL) 4147 4148 static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits) 4149 { 4150 /* use default */ 4151 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8)) 4152 return default_msg_enable_bits; 4153 if (debug_value == 0) /* no output */ 4154 return 0; 4155 /* set low N bits */ 4156 return (1U << debug_value) - 1; 4157 } 4158 4159 static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu) 4160 { 4161 spin_lock(&txq->_xmit_lock); 4162 /* Pairs with READ_ONCE() in __dev_queue_xmit() */ 4163 WRITE_ONCE(txq->xmit_lock_owner, cpu); 4164 } 4165 4166 static inline bool __netif_tx_acquire(struct netdev_queue *txq) 4167 { 4168 __acquire(&txq->_xmit_lock); 4169 return true; 4170 } 4171 4172 static inline void __netif_tx_release(struct netdev_queue *txq) 4173 { 4174 __release(&txq->_xmit_lock); 4175 } 4176 4177 static inline void __netif_tx_lock_bh(struct netdev_queue *txq) 4178 { 4179 spin_lock_bh(&txq->_xmit_lock); 4180 /* Pairs with READ_ONCE() in __dev_queue_xmit() */ 4181 WRITE_ONCE(txq->xmit_lock_owner, smp_processor_id()); 4182 } 4183 4184 static inline bool __netif_tx_trylock(struct netdev_queue *txq) 4185 { 4186 bool ok = spin_trylock(&txq->_xmit_lock); 4187 4188 if (likely(ok)) { 4189 /* Pairs with READ_ONCE() in __dev_queue_xmit() */ 4190 WRITE_ONCE(txq->xmit_lock_owner, smp_processor_id()); 4191 } 4192 return ok; 4193 } 4194 4195 static inline void __netif_tx_unlock(struct netdev_queue *txq) 4196 { 4197 /* Pairs with READ_ONCE() in __dev_queue_xmit() */ 4198 WRITE_ONCE(txq->xmit_lock_owner, -1); 4199 spin_unlock(&txq->_xmit_lock); 4200 } 4201 4202 static inline void __netif_tx_unlock_bh(struct netdev_queue *txq) 4203 { 4204 /* Pairs with READ_ONCE() in __dev_queue_xmit() */ 4205 WRITE_ONCE(txq->xmit_lock_owner, -1); 4206 spin_unlock_bh(&txq->_xmit_lock); 4207 } 4208 4209 /* 4210 * txq->trans_start can be read locklessly from dev_watchdog() 4211 */ 4212 static inline void txq_trans_update(struct netdev_queue *txq) 4213 { 4214 if (txq->xmit_lock_owner != -1) 4215 WRITE_ONCE(txq->trans_start, jiffies); 4216 } 4217 4218 static inline void txq_trans_cond_update(struct netdev_queue *txq) 4219 { 4220 unsigned long now = jiffies; 4221 4222 if (READ_ONCE(txq->trans_start) != now) 4223 WRITE_ONCE(txq->trans_start, now); 4224 } 4225 4226 /* legacy drivers only, netdev_start_xmit() sets txq->trans_start */ 4227 static inline void netif_trans_update(struct net_device *dev) 4228 { 4229 struct netdev_queue *txq = netdev_get_tx_queue(dev, 0); 4230 4231 txq_trans_cond_update(txq); 4232 } 4233 4234 /** 4235 * netif_tx_lock - grab network device transmit lock 4236 * @dev: network device 4237 * 4238 * Get network device transmit lock 4239 */ 4240 void netif_tx_lock(struct net_device *dev); 4241 4242 static inline void netif_tx_lock_bh(struct net_device *dev) 4243 { 4244 local_bh_disable(); 4245 netif_tx_lock(dev); 4246 } 4247 4248 void netif_tx_unlock(struct net_device *dev); 4249 4250 static inline void netif_tx_unlock_bh(struct net_device *dev) 4251 { 4252 netif_tx_unlock(dev); 4253 local_bh_enable(); 4254 } 4255 4256 #define HARD_TX_LOCK(dev, txq, cpu) { \ 4257 if ((dev->features & NETIF_F_LLTX) == 0) { \ 4258 __netif_tx_lock(txq, cpu); \ 4259 } else { \ 4260 __netif_tx_acquire(txq); \ 4261 } \ 4262 } 4263 4264 #define HARD_TX_TRYLOCK(dev, txq) \ 4265 (((dev->features & NETIF_F_LLTX) == 0) ? \ 4266 __netif_tx_trylock(txq) : \ 4267 __netif_tx_acquire(txq)) 4268 4269 #define HARD_TX_UNLOCK(dev, txq) { \ 4270 if ((dev->features & NETIF_F_LLTX) == 0) { \ 4271 __netif_tx_unlock(txq); \ 4272 } else { \ 4273 __netif_tx_release(txq); \ 4274 } \ 4275 } 4276 4277 static inline void netif_tx_disable(struct net_device *dev) 4278 { 4279 unsigned int i; 4280 int cpu; 4281 4282 local_bh_disable(); 4283 cpu = smp_processor_id(); 4284 spin_lock(&dev->tx_global_lock); 4285 for (i = 0; i < dev->num_tx_queues; i++) { 4286 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 4287 4288 __netif_tx_lock(txq, cpu); 4289 netif_tx_stop_queue(txq); 4290 __netif_tx_unlock(txq); 4291 } 4292 spin_unlock(&dev->tx_global_lock); 4293 local_bh_enable(); 4294 } 4295 4296 static inline void netif_addr_lock(struct net_device *dev) 4297 { 4298 unsigned char nest_level = 0; 4299 4300 #ifdef CONFIG_LOCKDEP 4301 nest_level = dev->nested_level; 4302 #endif 4303 spin_lock_nested(&dev->addr_list_lock, nest_level); 4304 } 4305 4306 static inline void netif_addr_lock_bh(struct net_device *dev) 4307 { 4308 unsigned char nest_level = 0; 4309 4310 #ifdef CONFIG_LOCKDEP 4311 nest_level = dev->nested_level; 4312 #endif 4313 local_bh_disable(); 4314 spin_lock_nested(&dev->addr_list_lock, nest_level); 4315 } 4316 4317 static inline void netif_addr_unlock(struct net_device *dev) 4318 { 4319 spin_unlock(&dev->addr_list_lock); 4320 } 4321 4322 static inline void netif_addr_unlock_bh(struct net_device *dev) 4323 { 4324 spin_unlock_bh(&dev->addr_list_lock); 4325 } 4326 4327 /* 4328 * dev_addrs walker. Should be used only for read access. Call with 4329 * rcu_read_lock held. 4330 */ 4331 #define for_each_dev_addr(dev, ha) \ 4332 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list) 4333 4334 /* These functions live elsewhere (drivers/net/net_init.c, but related) */ 4335 4336 void ether_setup(struct net_device *dev); 4337 4338 /* Support for loadable net-drivers */ 4339 struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name, 4340 unsigned char name_assign_type, 4341 void (*setup)(struct net_device *), 4342 unsigned int txqs, unsigned int rxqs); 4343 #define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \ 4344 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1) 4345 4346 #define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \ 4347 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \ 4348 count) 4349 4350 int register_netdev(struct net_device *dev); 4351 void unregister_netdev(struct net_device *dev); 4352 4353 int devm_register_netdev(struct device *dev, struct net_device *ndev); 4354 4355 /* General hardware address lists handling functions */ 4356 int __hw_addr_sync(struct netdev_hw_addr_list *to_list, 4357 struct netdev_hw_addr_list *from_list, int addr_len); 4358 void __hw_addr_unsync(struct netdev_hw_addr_list *to_list, 4359 struct netdev_hw_addr_list *from_list, int addr_len); 4360 int __hw_addr_sync_dev(struct netdev_hw_addr_list *list, 4361 struct net_device *dev, 4362 int (*sync)(struct net_device *, const unsigned char *), 4363 int (*unsync)(struct net_device *, 4364 const unsigned char *)); 4365 int __hw_addr_ref_sync_dev(struct netdev_hw_addr_list *list, 4366 struct net_device *dev, 4367 int (*sync)(struct net_device *, 4368 const unsigned char *, int), 4369 int (*unsync)(struct net_device *, 4370 const unsigned char *, int)); 4371 void __hw_addr_ref_unsync_dev(struct netdev_hw_addr_list *list, 4372 struct net_device *dev, 4373 int (*unsync)(struct net_device *, 4374 const unsigned char *, int)); 4375 void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list, 4376 struct net_device *dev, 4377 int (*unsync)(struct net_device *, 4378 const unsigned char *)); 4379 void __hw_addr_init(struct netdev_hw_addr_list *list); 4380 4381 /* Functions used for device addresses handling */ 4382 void dev_addr_mod(struct net_device *dev, unsigned int offset, 4383 const void *addr, size_t len); 4384 4385 static inline void 4386 __dev_addr_set(struct net_device *dev, const void *addr, size_t len) 4387 { 4388 dev_addr_mod(dev, 0, addr, len); 4389 } 4390 4391 static inline void dev_addr_set(struct net_device *dev, const u8 *addr) 4392 { 4393 __dev_addr_set(dev, addr, dev->addr_len); 4394 } 4395 4396 int dev_addr_add(struct net_device *dev, const unsigned char *addr, 4397 unsigned char addr_type); 4398 int dev_addr_del(struct net_device *dev, const unsigned char *addr, 4399 unsigned char addr_type); 4400 void dev_addr_flush(struct net_device *dev); 4401 int dev_addr_init(struct net_device *dev); 4402 void dev_addr_check(struct net_device *dev); 4403 4404 /* Functions used for unicast addresses handling */ 4405 int dev_uc_add(struct net_device *dev, const unsigned char *addr); 4406 int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr); 4407 int dev_uc_del(struct net_device *dev, const unsigned char *addr); 4408 int dev_uc_sync(struct net_device *to, struct net_device *from); 4409 int dev_uc_sync_multiple(struct net_device *to, struct net_device *from); 4410 void dev_uc_unsync(struct net_device *to, struct net_device *from); 4411 void dev_uc_flush(struct net_device *dev); 4412 void dev_uc_init(struct net_device *dev); 4413 4414 /** 4415 * __dev_uc_sync - Synchonize device's unicast list 4416 * @dev: device to sync 4417 * @sync: function to call if address should be added 4418 * @unsync: function to call if address should be removed 4419 * 4420 * Add newly added addresses to the interface, and release 4421 * addresses that have been deleted. 4422 */ 4423 static inline int __dev_uc_sync(struct net_device *dev, 4424 int (*sync)(struct net_device *, 4425 const unsigned char *), 4426 int (*unsync)(struct net_device *, 4427 const unsigned char *)) 4428 { 4429 return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync); 4430 } 4431 4432 /** 4433 * __dev_uc_unsync - Remove synchronized addresses from device 4434 * @dev: device to sync 4435 * @unsync: function to call if address should be removed 4436 * 4437 * Remove all addresses that were added to the device by dev_uc_sync(). 4438 */ 4439 static inline void __dev_uc_unsync(struct net_device *dev, 4440 int (*unsync)(struct net_device *, 4441 const unsigned char *)) 4442 { 4443 __hw_addr_unsync_dev(&dev->uc, dev, unsync); 4444 } 4445 4446 /* Functions used for multicast addresses handling */ 4447 int dev_mc_add(struct net_device *dev, const unsigned char *addr); 4448 int dev_mc_add_global(struct net_device *dev, const unsigned char *addr); 4449 int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr); 4450 int dev_mc_del(struct net_device *dev, const unsigned char *addr); 4451 int dev_mc_del_global(struct net_device *dev, const unsigned char *addr); 4452 int dev_mc_sync(struct net_device *to, struct net_device *from); 4453 int dev_mc_sync_multiple(struct net_device *to, struct net_device *from); 4454 void dev_mc_unsync(struct net_device *to, struct net_device *from); 4455 void dev_mc_flush(struct net_device *dev); 4456 void dev_mc_init(struct net_device *dev); 4457 4458 /** 4459 * __dev_mc_sync - Synchonize device's multicast list 4460 * @dev: device to sync 4461 * @sync: function to call if address should be added 4462 * @unsync: function to call if address should be removed 4463 * 4464 * Add newly added addresses to the interface, and release 4465 * addresses that have been deleted. 4466 */ 4467 static inline int __dev_mc_sync(struct net_device *dev, 4468 int (*sync)(struct net_device *, 4469 const unsigned char *), 4470 int (*unsync)(struct net_device *, 4471 const unsigned char *)) 4472 { 4473 return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync); 4474 } 4475 4476 /** 4477 * __dev_mc_unsync - Remove synchronized addresses from device 4478 * @dev: device to sync 4479 * @unsync: function to call if address should be removed 4480 * 4481 * Remove all addresses that were added to the device by dev_mc_sync(). 4482 */ 4483 static inline void __dev_mc_unsync(struct net_device *dev, 4484 int (*unsync)(struct net_device *, 4485 const unsigned char *)) 4486 { 4487 __hw_addr_unsync_dev(&dev->mc, dev, unsync); 4488 } 4489 4490 /* Functions used for secondary unicast and multicast support */ 4491 void dev_set_rx_mode(struct net_device *dev); 4492 void __dev_set_rx_mode(struct net_device *dev); 4493 int dev_set_promiscuity(struct net_device *dev, int inc); 4494 int dev_set_allmulti(struct net_device *dev, int inc); 4495 void netdev_state_change(struct net_device *dev); 4496 void __netdev_notify_peers(struct net_device *dev); 4497 void netdev_notify_peers(struct net_device *dev); 4498 void netdev_features_change(struct net_device *dev); 4499 /* Load a device via the kmod */ 4500 void dev_load(struct net *net, const char *name); 4501 struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev, 4502 struct rtnl_link_stats64 *storage); 4503 void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64, 4504 const struct net_device_stats *netdev_stats); 4505 void dev_fetch_sw_netstats(struct rtnl_link_stats64 *s, 4506 const struct pcpu_sw_netstats __percpu *netstats); 4507 void dev_get_tstats64(struct net_device *dev, struct rtnl_link_stats64 *s); 4508 4509 extern int netdev_max_backlog; 4510 extern int netdev_tstamp_prequeue; 4511 extern int netdev_unregister_timeout_secs; 4512 extern int weight_p; 4513 extern int dev_weight_rx_bias; 4514 extern int dev_weight_tx_bias; 4515 extern int dev_rx_weight; 4516 extern int dev_tx_weight; 4517 extern int gro_normal_batch; 4518 4519 enum { 4520 NESTED_SYNC_IMM_BIT, 4521 NESTED_SYNC_TODO_BIT, 4522 }; 4523 4524 #define __NESTED_SYNC_BIT(bit) ((u32)1 << (bit)) 4525 #define __NESTED_SYNC(name) __NESTED_SYNC_BIT(NESTED_SYNC_ ## name ## _BIT) 4526 4527 #define NESTED_SYNC_IMM __NESTED_SYNC(IMM) 4528 #define NESTED_SYNC_TODO __NESTED_SYNC(TODO) 4529 4530 struct netdev_nested_priv { 4531 unsigned char flags; 4532 void *data; 4533 }; 4534 4535 bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev); 4536 struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev, 4537 struct list_head **iter); 4538 4539 #ifdef CONFIG_LOCKDEP 4540 static LIST_HEAD(net_unlink_list); 4541 4542 static inline void net_unlink_todo(struct net_device *dev) 4543 { 4544 if (list_empty(&dev->unlink_list)) 4545 list_add_tail(&dev->unlink_list, &net_unlink_list); 4546 } 4547 #endif 4548 4549 /* iterate through upper list, must be called under RCU read lock */ 4550 #define netdev_for_each_upper_dev_rcu(dev, updev, iter) \ 4551 for (iter = &(dev)->adj_list.upper, \ 4552 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \ 4553 updev; \ 4554 updev = netdev_upper_get_next_dev_rcu(dev, &(iter))) 4555 4556 int netdev_walk_all_upper_dev_rcu(struct net_device *dev, 4557 int (*fn)(struct net_device *upper_dev, 4558 struct netdev_nested_priv *priv), 4559 struct netdev_nested_priv *priv); 4560 4561 bool netdev_has_upper_dev_all_rcu(struct net_device *dev, 4562 struct net_device *upper_dev); 4563 4564 bool netdev_has_any_upper_dev(struct net_device *dev); 4565 4566 void *netdev_lower_get_next_private(struct net_device *dev, 4567 struct list_head **iter); 4568 void *netdev_lower_get_next_private_rcu(struct net_device *dev, 4569 struct list_head **iter); 4570 4571 #define netdev_for_each_lower_private(dev, priv, iter) \ 4572 for (iter = (dev)->adj_list.lower.next, \ 4573 priv = netdev_lower_get_next_private(dev, &(iter)); \ 4574 priv; \ 4575 priv = netdev_lower_get_next_private(dev, &(iter))) 4576 4577 #define netdev_for_each_lower_private_rcu(dev, priv, iter) \ 4578 for (iter = &(dev)->adj_list.lower, \ 4579 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \ 4580 priv; \ 4581 priv = netdev_lower_get_next_private_rcu(dev, &(iter))) 4582 4583 void *netdev_lower_get_next(struct net_device *dev, 4584 struct list_head **iter); 4585 4586 #define netdev_for_each_lower_dev(dev, ldev, iter) \ 4587 for (iter = (dev)->adj_list.lower.next, \ 4588 ldev = netdev_lower_get_next(dev, &(iter)); \ 4589 ldev; \ 4590 ldev = netdev_lower_get_next(dev, &(iter))) 4591 4592 struct net_device *netdev_next_lower_dev_rcu(struct net_device *dev, 4593 struct list_head **iter); 4594 int netdev_walk_all_lower_dev(struct net_device *dev, 4595 int (*fn)(struct net_device *lower_dev, 4596 struct netdev_nested_priv *priv), 4597 struct netdev_nested_priv *priv); 4598 int netdev_walk_all_lower_dev_rcu(struct net_device *dev, 4599 int (*fn)(struct net_device *lower_dev, 4600 struct netdev_nested_priv *priv), 4601 struct netdev_nested_priv *priv); 4602 4603 void *netdev_adjacent_get_private(struct list_head *adj_list); 4604 void *netdev_lower_get_first_private_rcu(struct net_device *dev); 4605 struct net_device *netdev_master_upper_dev_get(struct net_device *dev); 4606 struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev); 4607 int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev, 4608 struct netlink_ext_ack *extack); 4609 int netdev_master_upper_dev_link(struct net_device *dev, 4610 struct net_device *upper_dev, 4611 void *upper_priv, void *upper_info, 4612 struct netlink_ext_ack *extack); 4613 void netdev_upper_dev_unlink(struct net_device *dev, 4614 struct net_device *upper_dev); 4615 int netdev_adjacent_change_prepare(struct net_device *old_dev, 4616 struct net_device *new_dev, 4617 struct net_device *dev, 4618 struct netlink_ext_ack *extack); 4619 void netdev_adjacent_change_commit(struct net_device *old_dev, 4620 struct net_device *new_dev, 4621 struct net_device *dev); 4622 void netdev_adjacent_change_abort(struct net_device *old_dev, 4623 struct net_device *new_dev, 4624 struct net_device *dev); 4625 void netdev_adjacent_rename_links(struct net_device *dev, char *oldname); 4626 void *netdev_lower_dev_get_private(struct net_device *dev, 4627 struct net_device *lower_dev); 4628 void netdev_lower_state_changed(struct net_device *lower_dev, 4629 void *lower_state_info); 4630 4631 /* RSS keys are 40 or 52 bytes long */ 4632 #define NETDEV_RSS_KEY_LEN 52 4633 extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN] __read_mostly; 4634 void netdev_rss_key_fill(void *buffer, size_t len); 4635 4636 int skb_checksum_help(struct sk_buff *skb); 4637 int skb_crc32c_csum_help(struct sk_buff *skb); 4638 int skb_csum_hwoffload_help(struct sk_buff *skb, 4639 const netdev_features_t features); 4640 4641 struct sk_buff *__skb_gso_segment(struct sk_buff *skb, 4642 netdev_features_t features, bool tx_path); 4643 struct sk_buff *skb_mac_gso_segment(struct sk_buff *skb, 4644 netdev_features_t features); 4645 4646 struct netdev_bonding_info { 4647 ifslave slave; 4648 ifbond master; 4649 }; 4650 4651 struct netdev_notifier_bonding_info { 4652 struct netdev_notifier_info info; /* must be first */ 4653 struct netdev_bonding_info bonding_info; 4654 }; 4655 4656 void netdev_bonding_info_change(struct net_device *dev, 4657 struct netdev_bonding_info *bonding_info); 4658 4659 #if IS_ENABLED(CONFIG_ETHTOOL_NETLINK) 4660 void ethtool_notify(struct net_device *dev, unsigned int cmd, const void *data); 4661 #else 4662 static inline void ethtool_notify(struct net_device *dev, unsigned int cmd, 4663 const void *data) 4664 { 4665 } 4666 #endif 4667 4668 static inline 4669 struct sk_buff *skb_gso_segment(struct sk_buff *skb, netdev_features_t features) 4670 { 4671 return __skb_gso_segment(skb, features, true); 4672 } 4673 __be16 skb_network_protocol(struct sk_buff *skb, int *depth); 4674 4675 static inline bool can_checksum_protocol(netdev_features_t features, 4676 __be16 protocol) 4677 { 4678 if (protocol == htons(ETH_P_FCOE)) 4679 return !!(features & NETIF_F_FCOE_CRC); 4680 4681 /* Assume this is an IP checksum (not SCTP CRC) */ 4682 4683 if (features & NETIF_F_HW_CSUM) { 4684 /* Can checksum everything */ 4685 return true; 4686 } 4687 4688 switch (protocol) { 4689 case htons(ETH_P_IP): 4690 return !!(features & NETIF_F_IP_CSUM); 4691 case htons(ETH_P_IPV6): 4692 return !!(features & NETIF_F_IPV6_CSUM); 4693 default: 4694 return false; 4695 } 4696 } 4697 4698 #ifdef CONFIG_BUG 4699 void netdev_rx_csum_fault(struct net_device *dev, struct sk_buff *skb); 4700 #else 4701 static inline void netdev_rx_csum_fault(struct net_device *dev, 4702 struct sk_buff *skb) 4703 { 4704 } 4705 #endif 4706 /* rx skb timestamps */ 4707 void net_enable_timestamp(void); 4708 void net_disable_timestamp(void); 4709 4710 #ifdef CONFIG_PROC_FS 4711 int __init dev_proc_init(void); 4712 #else 4713 #define dev_proc_init() 0 4714 #endif 4715 4716 static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops, 4717 struct sk_buff *skb, struct net_device *dev, 4718 bool more) 4719 { 4720 __this_cpu_write(softnet_data.xmit.more, more); 4721 return ops->ndo_start_xmit(skb, dev); 4722 } 4723 4724 static inline bool netdev_xmit_more(void) 4725 { 4726 return __this_cpu_read(softnet_data.xmit.more); 4727 } 4728 4729 static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev, 4730 struct netdev_queue *txq, bool more) 4731 { 4732 const struct net_device_ops *ops = dev->netdev_ops; 4733 netdev_tx_t rc; 4734 4735 rc = __netdev_start_xmit(ops, skb, dev, more); 4736 if (rc == NETDEV_TX_OK) 4737 txq_trans_update(txq); 4738 4739 return rc; 4740 } 4741 4742 int netdev_class_create_file_ns(const struct class_attribute *class_attr, 4743 const void *ns); 4744 void netdev_class_remove_file_ns(const struct class_attribute *class_attr, 4745 const void *ns); 4746 4747 extern const struct kobj_ns_type_operations net_ns_type_operations; 4748 4749 const char *netdev_drivername(const struct net_device *dev); 4750 4751 void linkwatch_run_queue(void); 4752 4753 static inline netdev_features_t netdev_intersect_features(netdev_features_t f1, 4754 netdev_features_t f2) 4755 { 4756 if ((f1 ^ f2) & NETIF_F_HW_CSUM) { 4757 if (f1 & NETIF_F_HW_CSUM) 4758 f1 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM); 4759 else 4760 f2 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM); 4761 } 4762 4763 return f1 & f2; 4764 } 4765 4766 static inline netdev_features_t netdev_get_wanted_features( 4767 struct net_device *dev) 4768 { 4769 return (dev->features & ~dev->hw_features) | dev->wanted_features; 4770 } 4771 netdev_features_t netdev_increment_features(netdev_features_t all, 4772 netdev_features_t one, netdev_features_t mask); 4773 4774 /* Allow TSO being used on stacked device : 4775 * Performing the GSO segmentation before last device 4776 * is a performance improvement. 4777 */ 4778 static inline netdev_features_t netdev_add_tso_features(netdev_features_t features, 4779 netdev_features_t mask) 4780 { 4781 return netdev_increment_features(features, NETIF_F_ALL_TSO, mask); 4782 } 4783 4784 int __netdev_update_features(struct net_device *dev); 4785 void netdev_update_features(struct net_device *dev); 4786 void netdev_change_features(struct net_device *dev); 4787 4788 void netif_stacked_transfer_operstate(const struct net_device *rootdev, 4789 struct net_device *dev); 4790 4791 netdev_features_t passthru_features_check(struct sk_buff *skb, 4792 struct net_device *dev, 4793 netdev_features_t features); 4794 netdev_features_t netif_skb_features(struct sk_buff *skb); 4795 4796 static inline bool net_gso_ok(netdev_features_t features, int gso_type) 4797 { 4798 netdev_features_t feature = (netdev_features_t)gso_type << NETIF_F_GSO_SHIFT; 4799 4800 /* check flags correspondence */ 4801 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT)); 4802 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT)); 4803 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT)); 4804 BUILD_BUG_ON(SKB_GSO_TCP_FIXEDID != (NETIF_F_TSO_MANGLEID >> NETIF_F_GSO_SHIFT)); 4805 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT)); 4806 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT)); 4807 BUILD_BUG_ON(SKB_GSO_GRE != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT)); 4808 BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT)); 4809 BUILD_BUG_ON(SKB_GSO_IPXIP4 != (NETIF_F_GSO_IPXIP4 >> NETIF_F_GSO_SHIFT)); 4810 BUILD_BUG_ON(SKB_GSO_IPXIP6 != (NETIF_F_GSO_IPXIP6 >> NETIF_F_GSO_SHIFT)); 4811 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT)); 4812 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT)); 4813 BUILD_BUG_ON(SKB_GSO_PARTIAL != (NETIF_F_GSO_PARTIAL >> NETIF_F_GSO_SHIFT)); 4814 BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT)); 4815 BUILD_BUG_ON(SKB_GSO_SCTP != (NETIF_F_GSO_SCTP >> NETIF_F_GSO_SHIFT)); 4816 BUILD_BUG_ON(SKB_GSO_ESP != (NETIF_F_GSO_ESP >> NETIF_F_GSO_SHIFT)); 4817 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_GSO_UDP >> NETIF_F_GSO_SHIFT)); 4818 BUILD_BUG_ON(SKB_GSO_UDP_L4 != (NETIF_F_GSO_UDP_L4 >> NETIF_F_GSO_SHIFT)); 4819 BUILD_BUG_ON(SKB_GSO_FRAGLIST != (NETIF_F_GSO_FRAGLIST >> NETIF_F_GSO_SHIFT)); 4820 4821 return (features & feature) == feature; 4822 } 4823 4824 static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features) 4825 { 4826 return net_gso_ok(features, skb_shinfo(skb)->gso_type) && 4827 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST)); 4828 } 4829 4830 static inline bool netif_needs_gso(struct sk_buff *skb, 4831 netdev_features_t features) 4832 { 4833 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) || 4834 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) && 4835 (skb->ip_summed != CHECKSUM_UNNECESSARY))); 4836 } 4837 4838 static inline void netif_set_gso_max_size(struct net_device *dev, 4839 unsigned int size) 4840 { 4841 /* dev->gso_max_size is read locklessly from sk_setup_caps() */ 4842 WRITE_ONCE(dev->gso_max_size, size); 4843 } 4844 4845 static inline void netif_set_gso_max_segs(struct net_device *dev, 4846 unsigned int segs) 4847 { 4848 /* dev->gso_max_segs is read locklessly from sk_setup_caps() */ 4849 WRITE_ONCE(dev->gso_max_segs, segs); 4850 } 4851 4852 static inline void netif_set_gro_max_size(struct net_device *dev, 4853 unsigned int size) 4854 { 4855 /* This pairs with the READ_ONCE() in skb_gro_receive() */ 4856 WRITE_ONCE(dev->gro_max_size, size); 4857 } 4858 4859 static inline void skb_gso_error_unwind(struct sk_buff *skb, __be16 protocol, 4860 int pulled_hlen, u16 mac_offset, 4861 int mac_len) 4862 { 4863 skb->protocol = protocol; 4864 skb->encapsulation = 1; 4865 skb_push(skb, pulled_hlen); 4866 skb_reset_transport_header(skb); 4867 skb->mac_header = mac_offset; 4868 skb->network_header = skb->mac_header + mac_len; 4869 skb->mac_len = mac_len; 4870 } 4871 4872 static inline bool netif_is_macsec(const struct net_device *dev) 4873 { 4874 return dev->priv_flags & IFF_MACSEC; 4875 } 4876 4877 static inline bool netif_is_macvlan(const struct net_device *dev) 4878 { 4879 return dev->priv_flags & IFF_MACVLAN; 4880 } 4881 4882 static inline bool netif_is_macvlan_port(const struct net_device *dev) 4883 { 4884 return dev->priv_flags & IFF_MACVLAN_PORT; 4885 } 4886 4887 static inline bool netif_is_bond_master(const struct net_device *dev) 4888 { 4889 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING; 4890 } 4891 4892 static inline bool netif_is_bond_slave(const struct net_device *dev) 4893 { 4894 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING; 4895 } 4896 4897 static inline bool netif_supports_nofcs(struct net_device *dev) 4898 { 4899 return dev->priv_flags & IFF_SUPP_NOFCS; 4900 } 4901 4902 static inline bool netif_has_l3_rx_handler(const struct net_device *dev) 4903 { 4904 return dev->priv_flags & IFF_L3MDEV_RX_HANDLER; 4905 } 4906 4907 static inline bool netif_is_l3_master(const struct net_device *dev) 4908 { 4909 return dev->priv_flags & IFF_L3MDEV_MASTER; 4910 } 4911 4912 static inline bool netif_is_l3_slave(const struct net_device *dev) 4913 { 4914 return dev->priv_flags & IFF_L3MDEV_SLAVE; 4915 } 4916 4917 static inline bool netif_is_bridge_master(const struct net_device *dev) 4918 { 4919 return dev->priv_flags & IFF_EBRIDGE; 4920 } 4921 4922 static inline bool netif_is_bridge_port(const struct net_device *dev) 4923 { 4924 return dev->priv_flags & IFF_BRIDGE_PORT; 4925 } 4926 4927 static inline bool netif_is_ovs_master(const struct net_device *dev) 4928 { 4929 return dev->priv_flags & IFF_OPENVSWITCH; 4930 } 4931 4932 static inline bool netif_is_ovs_port(const struct net_device *dev) 4933 { 4934 return dev->priv_flags & IFF_OVS_DATAPATH; 4935 } 4936 4937 static inline bool netif_is_any_bridge_port(const struct net_device *dev) 4938 { 4939 return netif_is_bridge_port(dev) || netif_is_ovs_port(dev); 4940 } 4941 4942 static inline bool netif_is_team_master(const struct net_device *dev) 4943 { 4944 return dev->priv_flags & IFF_TEAM; 4945 } 4946 4947 static inline bool netif_is_team_port(const struct net_device *dev) 4948 { 4949 return dev->priv_flags & IFF_TEAM_PORT; 4950 } 4951 4952 static inline bool netif_is_lag_master(const struct net_device *dev) 4953 { 4954 return netif_is_bond_master(dev) || netif_is_team_master(dev); 4955 } 4956 4957 static inline bool netif_is_lag_port(const struct net_device *dev) 4958 { 4959 return netif_is_bond_slave(dev) || netif_is_team_port(dev); 4960 } 4961 4962 static inline bool netif_is_rxfh_configured(const struct net_device *dev) 4963 { 4964 return dev->priv_flags & IFF_RXFH_CONFIGURED; 4965 } 4966 4967 static inline bool netif_is_failover(const struct net_device *dev) 4968 { 4969 return dev->priv_flags & IFF_FAILOVER; 4970 } 4971 4972 static inline bool netif_is_failover_slave(const struct net_device *dev) 4973 { 4974 return dev->priv_flags & IFF_FAILOVER_SLAVE; 4975 } 4976 4977 /* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */ 4978 static inline void netif_keep_dst(struct net_device *dev) 4979 { 4980 dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM); 4981 } 4982 4983 /* return true if dev can't cope with mtu frames that need vlan tag insertion */ 4984 static inline bool netif_reduces_vlan_mtu(struct net_device *dev) 4985 { 4986 /* TODO: reserve and use an additional IFF bit, if we get more users */ 4987 return netif_is_macsec(dev); 4988 } 4989 4990 extern struct pernet_operations __net_initdata loopback_net_ops; 4991 4992 /* Logging, debugging and troubleshooting/diagnostic helpers. */ 4993 4994 /* netdev_printk helpers, similar to dev_printk */ 4995 4996 static inline const char *netdev_name(const struct net_device *dev) 4997 { 4998 if (!dev->name[0] || strchr(dev->name, '%')) 4999 return "(unnamed net_device)"; 5000 return dev->name; 5001 } 5002 5003 static inline bool netdev_unregistering(const struct net_device *dev) 5004 { 5005 return dev->reg_state == NETREG_UNREGISTERING; 5006 } 5007 5008 static inline const char *netdev_reg_state(const struct net_device *dev) 5009 { 5010 switch (dev->reg_state) { 5011 case NETREG_UNINITIALIZED: return " (uninitialized)"; 5012 case NETREG_REGISTERED: return ""; 5013 case NETREG_UNREGISTERING: return " (unregistering)"; 5014 case NETREG_UNREGISTERED: return " (unregistered)"; 5015 case NETREG_RELEASED: return " (released)"; 5016 case NETREG_DUMMY: return " (dummy)"; 5017 } 5018 5019 WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, dev->reg_state); 5020 return " (unknown)"; 5021 } 5022 5023 __printf(3, 4) __cold 5024 void netdev_printk(const char *level, const struct net_device *dev, 5025 const char *format, ...); 5026 __printf(2, 3) __cold 5027 void netdev_emerg(const struct net_device *dev, const char *format, ...); 5028 __printf(2, 3) __cold 5029 void netdev_alert(const struct net_device *dev, const char *format, ...); 5030 __printf(2, 3) __cold 5031 void netdev_crit(const struct net_device *dev, const char *format, ...); 5032 __printf(2, 3) __cold 5033 void netdev_err(const struct net_device *dev, const char *format, ...); 5034 __printf(2, 3) __cold 5035 void netdev_warn(const struct net_device *dev, const char *format, ...); 5036 __printf(2, 3) __cold 5037 void netdev_notice(const struct net_device *dev, const char *format, ...); 5038 __printf(2, 3) __cold 5039 void netdev_info(const struct net_device *dev, const char *format, ...); 5040 5041 #define netdev_level_once(level, dev, fmt, ...) \ 5042 do { \ 5043 static bool __section(".data.once") __print_once; \ 5044 \ 5045 if (!__print_once) { \ 5046 __print_once = true; \ 5047 netdev_printk(level, dev, fmt, ##__VA_ARGS__); \ 5048 } \ 5049 } while (0) 5050 5051 #define netdev_emerg_once(dev, fmt, ...) \ 5052 netdev_level_once(KERN_EMERG, dev, fmt, ##__VA_ARGS__) 5053 #define netdev_alert_once(dev, fmt, ...) \ 5054 netdev_level_once(KERN_ALERT, dev, fmt, ##__VA_ARGS__) 5055 #define netdev_crit_once(dev, fmt, ...) \ 5056 netdev_level_once(KERN_CRIT, dev, fmt, ##__VA_ARGS__) 5057 #define netdev_err_once(dev, fmt, ...) \ 5058 netdev_level_once(KERN_ERR, dev, fmt, ##__VA_ARGS__) 5059 #define netdev_warn_once(dev, fmt, ...) \ 5060 netdev_level_once(KERN_WARNING, dev, fmt, ##__VA_ARGS__) 5061 #define netdev_notice_once(dev, fmt, ...) \ 5062 netdev_level_once(KERN_NOTICE, dev, fmt, ##__VA_ARGS__) 5063 #define netdev_info_once(dev, fmt, ...) \ 5064 netdev_level_once(KERN_INFO, dev, fmt, ##__VA_ARGS__) 5065 5066 #define MODULE_ALIAS_NETDEV(device) \ 5067 MODULE_ALIAS("netdev-" device) 5068 5069 #if defined(CONFIG_DYNAMIC_DEBUG) || \ 5070 (defined(CONFIG_DYNAMIC_DEBUG_CORE) && defined(DYNAMIC_DEBUG_MODULE)) 5071 #define netdev_dbg(__dev, format, args...) \ 5072 do { \ 5073 dynamic_netdev_dbg(__dev, format, ##args); \ 5074 } while (0) 5075 #elif defined(DEBUG) 5076 #define netdev_dbg(__dev, format, args...) \ 5077 netdev_printk(KERN_DEBUG, __dev, format, ##args) 5078 #else 5079 #define netdev_dbg(__dev, format, args...) \ 5080 ({ \ 5081 if (0) \ 5082 netdev_printk(KERN_DEBUG, __dev, format, ##args); \ 5083 }) 5084 #endif 5085 5086 #if defined(VERBOSE_DEBUG) 5087 #define netdev_vdbg netdev_dbg 5088 #else 5089 5090 #define netdev_vdbg(dev, format, args...) \ 5091 ({ \ 5092 if (0) \ 5093 netdev_printk(KERN_DEBUG, dev, format, ##args); \ 5094 0; \ 5095 }) 5096 #endif 5097 5098 /* 5099 * netdev_WARN() acts like dev_printk(), but with the key difference 5100 * of using a WARN/WARN_ON to get the message out, including the 5101 * file/line information and a backtrace. 5102 */ 5103 #define netdev_WARN(dev, format, args...) \ 5104 WARN(1, "netdevice: %s%s: " format, netdev_name(dev), \ 5105 netdev_reg_state(dev), ##args) 5106 5107 #define netdev_WARN_ONCE(dev, format, args...) \ 5108 WARN_ONCE(1, "netdevice: %s%s: " format, netdev_name(dev), \ 5109 netdev_reg_state(dev), ##args) 5110 5111 /* netif printk helpers, similar to netdev_printk */ 5112 5113 #define netif_printk(priv, type, level, dev, fmt, args...) \ 5114 do { \ 5115 if (netif_msg_##type(priv)) \ 5116 netdev_printk(level, (dev), fmt, ##args); \ 5117 } while (0) 5118 5119 #define netif_level(level, priv, type, dev, fmt, args...) \ 5120 do { \ 5121 if (netif_msg_##type(priv)) \ 5122 netdev_##level(dev, fmt, ##args); \ 5123 } while (0) 5124 5125 #define netif_emerg(priv, type, dev, fmt, args...) \ 5126 netif_level(emerg, priv, type, dev, fmt, ##args) 5127 #define netif_alert(priv, type, dev, fmt, args...) \ 5128 netif_level(alert, priv, type, dev, fmt, ##args) 5129 #define netif_crit(priv, type, dev, fmt, args...) \ 5130 netif_level(crit, priv, type, dev, fmt, ##args) 5131 #define netif_err(priv, type, dev, fmt, args...) \ 5132 netif_level(err, priv, type, dev, fmt, ##args) 5133 #define netif_warn(priv, type, dev, fmt, args...) \ 5134 netif_level(warn, priv, type, dev, fmt, ##args) 5135 #define netif_notice(priv, type, dev, fmt, args...) \ 5136 netif_level(notice, priv, type, dev, fmt, ##args) 5137 #define netif_info(priv, type, dev, fmt, args...) \ 5138 netif_level(info, priv, type, dev, fmt, ##args) 5139 5140 #if defined(CONFIG_DYNAMIC_DEBUG) || \ 5141 (defined(CONFIG_DYNAMIC_DEBUG_CORE) && defined(DYNAMIC_DEBUG_MODULE)) 5142 #define netif_dbg(priv, type, netdev, format, args...) \ 5143 do { \ 5144 if (netif_msg_##type(priv)) \ 5145 dynamic_netdev_dbg(netdev, format, ##args); \ 5146 } while (0) 5147 #elif defined(DEBUG) 5148 #define netif_dbg(priv, type, dev, format, args...) \ 5149 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args) 5150 #else 5151 #define netif_dbg(priv, type, dev, format, args...) \ 5152 ({ \ 5153 if (0) \ 5154 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \ 5155 0; \ 5156 }) 5157 #endif 5158 5159 /* if @cond then downgrade to debug, else print at @level */ 5160 #define netif_cond_dbg(priv, type, netdev, cond, level, fmt, args...) \ 5161 do { \ 5162 if (cond) \ 5163 netif_dbg(priv, type, netdev, fmt, ##args); \ 5164 else \ 5165 netif_ ## level(priv, type, netdev, fmt, ##args); \ 5166 } while (0) 5167 5168 #if defined(VERBOSE_DEBUG) 5169 #define netif_vdbg netif_dbg 5170 #else 5171 #define netif_vdbg(priv, type, dev, format, args...) \ 5172 ({ \ 5173 if (0) \ 5174 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \ 5175 0; \ 5176 }) 5177 #endif 5178 5179 /* 5180 * The list of packet types we will receive (as opposed to discard) 5181 * and the routines to invoke. 5182 * 5183 * Why 16. Because with 16 the only overlap we get on a hash of the 5184 * low nibble of the protocol value is RARP/SNAP/X.25. 5185 * 5186 * 0800 IP 5187 * 0001 802.3 5188 * 0002 AX.25 5189 * 0004 802.2 5190 * 8035 RARP 5191 * 0005 SNAP 5192 * 0805 X.25 5193 * 0806 ARP 5194 * 8137 IPX 5195 * 0009 Localtalk 5196 * 86DD IPv6 5197 */ 5198 #define PTYPE_HASH_SIZE (16) 5199 #define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1) 5200 5201 extern struct list_head ptype_all __read_mostly; 5202 extern struct list_head ptype_base[PTYPE_HASH_SIZE] __read_mostly; 5203 5204 extern struct net_device *blackhole_netdev; 5205 5206 #endif /* _LINUX_NETDEVICE_H */ 5207