1 /* SPDX-License-Identifier: GPL-2.0-or-later */ 2 /* 3 * INET An implementation of the TCP/IP protocol suite for the LINUX 4 * operating system. INET is implemented using the BSD Socket 5 * interface as the means of communication with the user level. 6 * 7 * Definitions for the Interfaces handler. 8 * 9 * Version: @(#)dev.h 1.0.10 08/12/93 10 * 11 * Authors: Ross Biro 12 * Fred N. van Kempen, <[email protected]> 13 * Corey Minyard <[email protected]> 14 * Donald J. Becker, <[email protected]> 15 * Alan Cox, <[email protected]> 16 * Bjorn Ekwall. <[email protected]> 17 * Pekka Riikonen <[email protected]> 18 * 19 * Moved to /usr/include/linux for NET3 20 */ 21 #ifndef _LINUX_NETDEVICE_H 22 #define _LINUX_NETDEVICE_H 23 24 #include <linux/timer.h> 25 #include <linux/bug.h> 26 #include <linux/delay.h> 27 #include <linux/atomic.h> 28 #include <linux/prefetch.h> 29 #include <asm/cache.h> 30 #include <asm/byteorder.h> 31 32 #include <linux/percpu.h> 33 #include <linux/rculist.h> 34 #include <linux/workqueue.h> 35 #include <linux/dynamic_queue_limits.h> 36 37 #include <net/net_namespace.h> 38 #ifdef CONFIG_DCB 39 #include <net/dcbnl.h> 40 #endif 41 #include <net/netprio_cgroup.h> 42 #include <net/xdp.h> 43 44 #include <linux/netdev_features.h> 45 #include <linux/neighbour.h> 46 #include <uapi/linux/netdevice.h> 47 #include <uapi/linux/if_bonding.h> 48 #include <uapi/linux/pkt_cls.h> 49 #include <linux/hashtable.h> 50 #include <linux/rbtree.h> 51 #include <net/net_trackers.h> 52 53 struct netpoll_info; 54 struct device; 55 struct ethtool_ops; 56 struct phy_device; 57 struct dsa_port; 58 struct ip_tunnel_parm; 59 struct macsec_context; 60 struct macsec_ops; 61 62 struct sfp_bus; 63 /* 802.11 specific */ 64 struct wireless_dev; 65 /* 802.15.4 specific */ 66 struct wpan_dev; 67 struct mpls_dev; 68 /* UDP Tunnel offloads */ 69 struct udp_tunnel_info; 70 struct udp_tunnel_nic_info; 71 struct udp_tunnel_nic; 72 struct bpf_prog; 73 struct xdp_buff; 74 75 void synchronize_net(void); 76 void netdev_set_default_ethtool_ops(struct net_device *dev, 77 const struct ethtool_ops *ops); 78 79 /* Backlog congestion levels */ 80 #define NET_RX_SUCCESS 0 /* keep 'em coming, baby */ 81 #define NET_RX_DROP 1 /* packet dropped */ 82 83 #define MAX_NEST_DEV 8 84 85 /* 86 * Transmit return codes: transmit return codes originate from three different 87 * namespaces: 88 * 89 * - qdisc return codes 90 * - driver transmit return codes 91 * - errno values 92 * 93 * Drivers are allowed to return any one of those in their hard_start_xmit() 94 * function. Real network devices commonly used with qdiscs should only return 95 * the driver transmit return codes though - when qdiscs are used, the actual 96 * transmission happens asynchronously, so the value is not propagated to 97 * higher layers. Virtual network devices transmit synchronously; in this case 98 * the driver transmit return codes are consumed by dev_queue_xmit(), and all 99 * others are propagated to higher layers. 100 */ 101 102 /* qdisc ->enqueue() return codes. */ 103 #define NET_XMIT_SUCCESS 0x00 104 #define NET_XMIT_DROP 0x01 /* skb dropped */ 105 #define NET_XMIT_CN 0x02 /* congestion notification */ 106 #define NET_XMIT_MASK 0x0f /* qdisc flags in net/sch_generic.h */ 107 108 /* NET_XMIT_CN is special. It does not guarantee that this packet is lost. It 109 * indicates that the device will soon be dropping packets, or already drops 110 * some packets of the same priority; prompting us to send less aggressively. */ 111 #define net_xmit_eval(e) ((e) == NET_XMIT_CN ? 0 : (e)) 112 #define net_xmit_errno(e) ((e) != NET_XMIT_CN ? -ENOBUFS : 0) 113 114 /* Driver transmit return codes */ 115 #define NETDEV_TX_MASK 0xf0 116 117 enum netdev_tx { 118 __NETDEV_TX_MIN = INT_MIN, /* make sure enum is signed */ 119 NETDEV_TX_OK = 0x00, /* driver took care of packet */ 120 NETDEV_TX_BUSY = 0x10, /* driver tx path was busy*/ 121 }; 122 typedef enum netdev_tx netdev_tx_t; 123 124 /* 125 * Current order: NETDEV_TX_MASK > NET_XMIT_MASK >= 0 is significant; 126 * hard_start_xmit() return < NET_XMIT_MASK means skb was consumed. 127 */ 128 static inline bool dev_xmit_complete(int rc) 129 { 130 /* 131 * Positive cases with an skb consumed by a driver: 132 * - successful transmission (rc == NETDEV_TX_OK) 133 * - error while transmitting (rc < 0) 134 * - error while queueing to a different device (rc & NET_XMIT_MASK) 135 */ 136 if (likely(rc < NET_XMIT_MASK)) 137 return true; 138 139 return false; 140 } 141 142 /* 143 * Compute the worst-case header length according to the protocols 144 * used. 145 */ 146 147 #if defined(CONFIG_HYPERV_NET) 148 # define LL_MAX_HEADER 128 149 #elif defined(CONFIG_WLAN) || IS_ENABLED(CONFIG_AX25) 150 # if defined(CONFIG_MAC80211_MESH) 151 # define LL_MAX_HEADER 128 152 # else 153 # define LL_MAX_HEADER 96 154 # endif 155 #else 156 # define LL_MAX_HEADER 32 157 #endif 158 159 #if !IS_ENABLED(CONFIG_NET_IPIP) && !IS_ENABLED(CONFIG_NET_IPGRE) && \ 160 !IS_ENABLED(CONFIG_IPV6_SIT) && !IS_ENABLED(CONFIG_IPV6_TUNNEL) 161 #define MAX_HEADER LL_MAX_HEADER 162 #else 163 #define MAX_HEADER (LL_MAX_HEADER + 48) 164 #endif 165 166 /* 167 * Old network device statistics. Fields are native words 168 * (unsigned long) so they can be read and written atomically. 169 */ 170 171 struct net_device_stats { 172 unsigned long rx_packets; 173 unsigned long tx_packets; 174 unsigned long rx_bytes; 175 unsigned long tx_bytes; 176 unsigned long rx_errors; 177 unsigned long tx_errors; 178 unsigned long rx_dropped; 179 unsigned long tx_dropped; 180 unsigned long multicast; 181 unsigned long collisions; 182 unsigned long rx_length_errors; 183 unsigned long rx_over_errors; 184 unsigned long rx_crc_errors; 185 unsigned long rx_frame_errors; 186 unsigned long rx_fifo_errors; 187 unsigned long rx_missed_errors; 188 unsigned long tx_aborted_errors; 189 unsigned long tx_carrier_errors; 190 unsigned long tx_fifo_errors; 191 unsigned long tx_heartbeat_errors; 192 unsigned long tx_window_errors; 193 unsigned long rx_compressed; 194 unsigned long tx_compressed; 195 }; 196 197 198 #include <linux/cache.h> 199 #include <linux/skbuff.h> 200 201 #ifdef CONFIG_RPS 202 #include <linux/static_key.h> 203 extern struct static_key_false rps_needed; 204 extern struct static_key_false rfs_needed; 205 #endif 206 207 struct neighbour; 208 struct neigh_parms; 209 struct sk_buff; 210 211 struct netdev_hw_addr { 212 struct list_head list; 213 struct rb_node node; 214 unsigned char addr[MAX_ADDR_LEN]; 215 unsigned char type; 216 #define NETDEV_HW_ADDR_T_LAN 1 217 #define NETDEV_HW_ADDR_T_SAN 2 218 #define NETDEV_HW_ADDR_T_UNICAST 3 219 #define NETDEV_HW_ADDR_T_MULTICAST 4 220 bool global_use; 221 int sync_cnt; 222 int refcount; 223 int synced; 224 struct rcu_head rcu_head; 225 }; 226 227 struct netdev_hw_addr_list { 228 struct list_head list; 229 int count; 230 231 /* Auxiliary tree for faster lookup on addition and deletion */ 232 struct rb_root tree; 233 }; 234 235 #define netdev_hw_addr_list_count(l) ((l)->count) 236 #define netdev_hw_addr_list_empty(l) (netdev_hw_addr_list_count(l) == 0) 237 #define netdev_hw_addr_list_for_each(ha, l) \ 238 list_for_each_entry(ha, &(l)->list, list) 239 240 #define netdev_uc_count(dev) netdev_hw_addr_list_count(&(dev)->uc) 241 #define netdev_uc_empty(dev) netdev_hw_addr_list_empty(&(dev)->uc) 242 #define netdev_for_each_uc_addr(ha, dev) \ 243 netdev_hw_addr_list_for_each(ha, &(dev)->uc) 244 245 #define netdev_mc_count(dev) netdev_hw_addr_list_count(&(dev)->mc) 246 #define netdev_mc_empty(dev) netdev_hw_addr_list_empty(&(dev)->mc) 247 #define netdev_for_each_mc_addr(ha, dev) \ 248 netdev_hw_addr_list_for_each(ha, &(dev)->mc) 249 250 struct hh_cache { 251 unsigned int hh_len; 252 seqlock_t hh_lock; 253 254 /* cached hardware header; allow for machine alignment needs. */ 255 #define HH_DATA_MOD 16 256 #define HH_DATA_OFF(__len) \ 257 (HH_DATA_MOD - (((__len - 1) & (HH_DATA_MOD - 1)) + 1)) 258 #define HH_DATA_ALIGN(__len) \ 259 (((__len)+(HH_DATA_MOD-1))&~(HH_DATA_MOD - 1)) 260 unsigned long hh_data[HH_DATA_ALIGN(LL_MAX_HEADER) / sizeof(long)]; 261 }; 262 263 /* Reserve HH_DATA_MOD byte-aligned hard_header_len, but at least that much. 264 * Alternative is: 265 * dev->hard_header_len ? (dev->hard_header_len + 266 * (HH_DATA_MOD - 1)) & ~(HH_DATA_MOD - 1) : 0 267 * 268 * We could use other alignment values, but we must maintain the 269 * relationship HH alignment <= LL alignment. 270 */ 271 #define LL_RESERVED_SPACE(dev) \ 272 ((((dev)->hard_header_len+(dev)->needed_headroom)&~(HH_DATA_MOD - 1)) + HH_DATA_MOD) 273 #define LL_RESERVED_SPACE_EXTRA(dev,extra) \ 274 ((((dev)->hard_header_len+(dev)->needed_headroom+(extra))&~(HH_DATA_MOD - 1)) + HH_DATA_MOD) 275 276 struct header_ops { 277 int (*create) (struct sk_buff *skb, struct net_device *dev, 278 unsigned short type, const void *daddr, 279 const void *saddr, unsigned int len); 280 int (*parse)(const struct sk_buff *skb, unsigned char *haddr); 281 int (*cache)(const struct neighbour *neigh, struct hh_cache *hh, __be16 type); 282 void (*cache_update)(struct hh_cache *hh, 283 const struct net_device *dev, 284 const unsigned char *haddr); 285 bool (*validate)(const char *ll_header, unsigned int len); 286 __be16 (*parse_protocol)(const struct sk_buff *skb); 287 }; 288 289 /* These flag bits are private to the generic network queueing 290 * layer; they may not be explicitly referenced by any other 291 * code. 292 */ 293 294 enum netdev_state_t { 295 __LINK_STATE_START, 296 __LINK_STATE_PRESENT, 297 __LINK_STATE_NOCARRIER, 298 __LINK_STATE_LINKWATCH_PENDING, 299 __LINK_STATE_DORMANT, 300 __LINK_STATE_TESTING, 301 }; 302 303 struct gro_list { 304 struct list_head list; 305 int count; 306 }; 307 308 /* 309 * size of gro hash buckets, must less than bit number of 310 * napi_struct::gro_bitmask 311 */ 312 #define GRO_HASH_BUCKETS 8 313 314 /* 315 * Structure for NAPI scheduling similar to tasklet but with weighting 316 */ 317 struct napi_struct { 318 /* The poll_list must only be managed by the entity which 319 * changes the state of the NAPI_STATE_SCHED bit. This means 320 * whoever atomically sets that bit can add this napi_struct 321 * to the per-CPU poll_list, and whoever clears that bit 322 * can remove from the list right before clearing the bit. 323 */ 324 struct list_head poll_list; 325 326 unsigned long state; 327 int weight; 328 int defer_hard_irqs_count; 329 unsigned long gro_bitmask; 330 int (*poll)(struct napi_struct *, int); 331 #ifdef CONFIG_NETPOLL 332 int poll_owner; 333 #endif 334 struct net_device *dev; 335 struct gro_list gro_hash[GRO_HASH_BUCKETS]; 336 struct sk_buff *skb; 337 struct list_head rx_list; /* Pending GRO_NORMAL skbs */ 338 int rx_count; /* length of rx_list */ 339 struct hrtimer timer; 340 struct list_head dev_list; 341 struct hlist_node napi_hash_node; 342 unsigned int napi_id; 343 struct task_struct *thread; 344 }; 345 346 enum { 347 NAPI_STATE_SCHED, /* Poll is scheduled */ 348 NAPI_STATE_MISSED, /* reschedule a napi */ 349 NAPI_STATE_DISABLE, /* Disable pending */ 350 NAPI_STATE_NPSVC, /* Netpoll - don't dequeue from poll_list */ 351 NAPI_STATE_LISTED, /* NAPI added to system lists */ 352 NAPI_STATE_NO_BUSY_POLL, /* Do not add in napi_hash, no busy polling */ 353 NAPI_STATE_IN_BUSY_POLL, /* sk_busy_loop() owns this NAPI */ 354 NAPI_STATE_PREFER_BUSY_POLL, /* prefer busy-polling over softirq processing*/ 355 NAPI_STATE_THREADED, /* The poll is performed inside its own thread*/ 356 NAPI_STATE_SCHED_THREADED, /* Napi is currently scheduled in threaded mode */ 357 }; 358 359 enum { 360 NAPIF_STATE_SCHED = BIT(NAPI_STATE_SCHED), 361 NAPIF_STATE_MISSED = BIT(NAPI_STATE_MISSED), 362 NAPIF_STATE_DISABLE = BIT(NAPI_STATE_DISABLE), 363 NAPIF_STATE_NPSVC = BIT(NAPI_STATE_NPSVC), 364 NAPIF_STATE_LISTED = BIT(NAPI_STATE_LISTED), 365 NAPIF_STATE_NO_BUSY_POLL = BIT(NAPI_STATE_NO_BUSY_POLL), 366 NAPIF_STATE_IN_BUSY_POLL = BIT(NAPI_STATE_IN_BUSY_POLL), 367 NAPIF_STATE_PREFER_BUSY_POLL = BIT(NAPI_STATE_PREFER_BUSY_POLL), 368 NAPIF_STATE_THREADED = BIT(NAPI_STATE_THREADED), 369 NAPIF_STATE_SCHED_THREADED = BIT(NAPI_STATE_SCHED_THREADED), 370 }; 371 372 enum gro_result { 373 GRO_MERGED, 374 GRO_MERGED_FREE, 375 GRO_HELD, 376 GRO_NORMAL, 377 GRO_CONSUMED, 378 }; 379 typedef enum gro_result gro_result_t; 380 381 /* 382 * enum rx_handler_result - Possible return values for rx_handlers. 383 * @RX_HANDLER_CONSUMED: skb was consumed by rx_handler, do not process it 384 * further. 385 * @RX_HANDLER_ANOTHER: Do another round in receive path. This is indicated in 386 * case skb->dev was changed by rx_handler. 387 * @RX_HANDLER_EXACT: Force exact delivery, no wildcard. 388 * @RX_HANDLER_PASS: Do nothing, pass the skb as if no rx_handler was called. 389 * 390 * rx_handlers are functions called from inside __netif_receive_skb(), to do 391 * special processing of the skb, prior to delivery to protocol handlers. 392 * 393 * Currently, a net_device can only have a single rx_handler registered. Trying 394 * to register a second rx_handler will return -EBUSY. 395 * 396 * To register a rx_handler on a net_device, use netdev_rx_handler_register(). 397 * To unregister a rx_handler on a net_device, use 398 * netdev_rx_handler_unregister(). 399 * 400 * Upon return, rx_handler is expected to tell __netif_receive_skb() what to 401 * do with the skb. 402 * 403 * If the rx_handler consumed the skb in some way, it should return 404 * RX_HANDLER_CONSUMED. This is appropriate when the rx_handler arranged for 405 * the skb to be delivered in some other way. 406 * 407 * If the rx_handler changed skb->dev, to divert the skb to another 408 * net_device, it should return RX_HANDLER_ANOTHER. The rx_handler for the 409 * new device will be called if it exists. 410 * 411 * If the rx_handler decides the skb should be ignored, it should return 412 * RX_HANDLER_EXACT. The skb will only be delivered to protocol handlers that 413 * are registered on exact device (ptype->dev == skb->dev). 414 * 415 * If the rx_handler didn't change skb->dev, but wants the skb to be normally 416 * delivered, it should return RX_HANDLER_PASS. 417 * 418 * A device without a registered rx_handler will behave as if rx_handler 419 * returned RX_HANDLER_PASS. 420 */ 421 422 enum rx_handler_result { 423 RX_HANDLER_CONSUMED, 424 RX_HANDLER_ANOTHER, 425 RX_HANDLER_EXACT, 426 RX_HANDLER_PASS, 427 }; 428 typedef enum rx_handler_result rx_handler_result_t; 429 typedef rx_handler_result_t rx_handler_func_t(struct sk_buff **pskb); 430 431 void __napi_schedule(struct napi_struct *n); 432 void __napi_schedule_irqoff(struct napi_struct *n); 433 434 static inline bool napi_disable_pending(struct napi_struct *n) 435 { 436 return test_bit(NAPI_STATE_DISABLE, &n->state); 437 } 438 439 static inline bool napi_prefer_busy_poll(struct napi_struct *n) 440 { 441 return test_bit(NAPI_STATE_PREFER_BUSY_POLL, &n->state); 442 } 443 444 bool napi_schedule_prep(struct napi_struct *n); 445 446 /** 447 * napi_schedule - schedule NAPI poll 448 * @n: NAPI context 449 * 450 * Schedule NAPI poll routine to be called if it is not already 451 * running. 452 */ 453 static inline void napi_schedule(struct napi_struct *n) 454 { 455 if (napi_schedule_prep(n)) 456 __napi_schedule(n); 457 } 458 459 /** 460 * napi_schedule_irqoff - schedule NAPI poll 461 * @n: NAPI context 462 * 463 * Variant of napi_schedule(), assuming hard irqs are masked. 464 */ 465 static inline void napi_schedule_irqoff(struct napi_struct *n) 466 { 467 if (napi_schedule_prep(n)) 468 __napi_schedule_irqoff(n); 469 } 470 471 /* Try to reschedule poll. Called by dev->poll() after napi_complete(). */ 472 static inline bool napi_reschedule(struct napi_struct *napi) 473 { 474 if (napi_schedule_prep(napi)) { 475 __napi_schedule(napi); 476 return true; 477 } 478 return false; 479 } 480 481 bool napi_complete_done(struct napi_struct *n, int work_done); 482 /** 483 * napi_complete - NAPI processing complete 484 * @n: NAPI context 485 * 486 * Mark NAPI processing as complete. 487 * Consider using napi_complete_done() instead. 488 * Return false if device should avoid rearming interrupts. 489 */ 490 static inline bool napi_complete(struct napi_struct *n) 491 { 492 return napi_complete_done(n, 0); 493 } 494 495 int dev_set_threaded(struct net_device *dev, bool threaded); 496 497 /** 498 * napi_disable - prevent NAPI from scheduling 499 * @n: NAPI context 500 * 501 * Stop NAPI from being scheduled on this context. 502 * Waits till any outstanding processing completes. 503 */ 504 void napi_disable(struct napi_struct *n); 505 506 void napi_enable(struct napi_struct *n); 507 508 /** 509 * napi_synchronize - wait until NAPI is not running 510 * @n: NAPI context 511 * 512 * Wait until NAPI is done being scheduled on this context. 513 * Waits till any outstanding processing completes but 514 * does not disable future activations. 515 */ 516 static inline void napi_synchronize(const struct napi_struct *n) 517 { 518 if (IS_ENABLED(CONFIG_SMP)) 519 while (test_bit(NAPI_STATE_SCHED, &n->state)) 520 msleep(1); 521 else 522 barrier(); 523 } 524 525 /** 526 * napi_if_scheduled_mark_missed - if napi is running, set the 527 * NAPIF_STATE_MISSED 528 * @n: NAPI context 529 * 530 * If napi is running, set the NAPIF_STATE_MISSED, and return true if 531 * NAPI is scheduled. 532 **/ 533 static inline bool napi_if_scheduled_mark_missed(struct napi_struct *n) 534 { 535 unsigned long val, new; 536 537 do { 538 val = READ_ONCE(n->state); 539 if (val & NAPIF_STATE_DISABLE) 540 return true; 541 542 if (!(val & NAPIF_STATE_SCHED)) 543 return false; 544 545 new = val | NAPIF_STATE_MISSED; 546 } while (cmpxchg(&n->state, val, new) != val); 547 548 return true; 549 } 550 551 enum netdev_queue_state_t { 552 __QUEUE_STATE_DRV_XOFF, 553 __QUEUE_STATE_STACK_XOFF, 554 __QUEUE_STATE_FROZEN, 555 }; 556 557 #define QUEUE_STATE_DRV_XOFF (1 << __QUEUE_STATE_DRV_XOFF) 558 #define QUEUE_STATE_STACK_XOFF (1 << __QUEUE_STATE_STACK_XOFF) 559 #define QUEUE_STATE_FROZEN (1 << __QUEUE_STATE_FROZEN) 560 561 #define QUEUE_STATE_ANY_XOFF (QUEUE_STATE_DRV_XOFF | QUEUE_STATE_STACK_XOFF) 562 #define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \ 563 QUEUE_STATE_FROZEN) 564 #define QUEUE_STATE_DRV_XOFF_OR_FROZEN (QUEUE_STATE_DRV_XOFF | \ 565 QUEUE_STATE_FROZEN) 566 567 /* 568 * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue. The 569 * netif_tx_* functions below are used to manipulate this flag. The 570 * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit 571 * queue independently. The netif_xmit_*stopped functions below are called 572 * to check if the queue has been stopped by the driver or stack (either 573 * of the XOFF bits are set in the state). Drivers should not need to call 574 * netif_xmit*stopped functions, they should only be using netif_tx_*. 575 */ 576 577 struct netdev_queue { 578 /* 579 * read-mostly part 580 */ 581 struct net_device *dev; 582 netdevice_tracker dev_tracker; 583 584 struct Qdisc __rcu *qdisc; 585 struct Qdisc *qdisc_sleeping; 586 #ifdef CONFIG_SYSFS 587 struct kobject kobj; 588 #endif 589 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA) 590 int numa_node; 591 #endif 592 unsigned long tx_maxrate; 593 /* 594 * Number of TX timeouts for this queue 595 * (/sys/class/net/DEV/Q/trans_timeout) 596 */ 597 atomic_long_t trans_timeout; 598 599 /* Subordinate device that the queue has been assigned to */ 600 struct net_device *sb_dev; 601 #ifdef CONFIG_XDP_SOCKETS 602 struct xsk_buff_pool *pool; 603 #endif 604 /* 605 * write-mostly part 606 */ 607 spinlock_t _xmit_lock ____cacheline_aligned_in_smp; 608 int xmit_lock_owner; 609 /* 610 * Time (in jiffies) of last Tx 611 */ 612 unsigned long trans_start; 613 614 unsigned long state; 615 616 #ifdef CONFIG_BQL 617 struct dql dql; 618 #endif 619 } ____cacheline_aligned_in_smp; 620 621 extern int sysctl_fb_tunnels_only_for_init_net; 622 extern int sysctl_devconf_inherit_init_net; 623 624 /* 625 * sysctl_fb_tunnels_only_for_init_net == 0 : For all netns 626 * == 1 : For initns only 627 * == 2 : For none. 628 */ 629 static inline bool net_has_fallback_tunnels(const struct net *net) 630 { 631 return !IS_ENABLED(CONFIG_SYSCTL) || 632 !sysctl_fb_tunnels_only_for_init_net || 633 (net == &init_net && sysctl_fb_tunnels_only_for_init_net == 1); 634 } 635 636 static inline int netdev_queue_numa_node_read(const struct netdev_queue *q) 637 { 638 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA) 639 return q->numa_node; 640 #else 641 return NUMA_NO_NODE; 642 #endif 643 } 644 645 static inline void netdev_queue_numa_node_write(struct netdev_queue *q, int node) 646 { 647 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA) 648 q->numa_node = node; 649 #endif 650 } 651 652 #ifdef CONFIG_RPS 653 /* 654 * This structure holds an RPS map which can be of variable length. The 655 * map is an array of CPUs. 656 */ 657 struct rps_map { 658 unsigned int len; 659 struct rcu_head rcu; 660 u16 cpus[]; 661 }; 662 #define RPS_MAP_SIZE(_num) (sizeof(struct rps_map) + ((_num) * sizeof(u16))) 663 664 /* 665 * The rps_dev_flow structure contains the mapping of a flow to a CPU, the 666 * tail pointer for that CPU's input queue at the time of last enqueue, and 667 * a hardware filter index. 668 */ 669 struct rps_dev_flow { 670 u16 cpu; 671 u16 filter; 672 unsigned int last_qtail; 673 }; 674 #define RPS_NO_FILTER 0xffff 675 676 /* 677 * The rps_dev_flow_table structure contains a table of flow mappings. 678 */ 679 struct rps_dev_flow_table { 680 unsigned int mask; 681 struct rcu_head rcu; 682 struct rps_dev_flow flows[]; 683 }; 684 #define RPS_DEV_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_dev_flow_table) + \ 685 ((_num) * sizeof(struct rps_dev_flow))) 686 687 /* 688 * The rps_sock_flow_table contains mappings of flows to the last CPU 689 * on which they were processed by the application (set in recvmsg). 690 * Each entry is a 32bit value. Upper part is the high-order bits 691 * of flow hash, lower part is CPU number. 692 * rps_cpu_mask is used to partition the space, depending on number of 693 * possible CPUs : rps_cpu_mask = roundup_pow_of_two(nr_cpu_ids) - 1 694 * For example, if 64 CPUs are possible, rps_cpu_mask = 0x3f, 695 * meaning we use 32-6=26 bits for the hash. 696 */ 697 struct rps_sock_flow_table { 698 u32 mask; 699 700 u32 ents[] ____cacheline_aligned_in_smp; 701 }; 702 #define RPS_SOCK_FLOW_TABLE_SIZE(_num) (offsetof(struct rps_sock_flow_table, ents[_num])) 703 704 #define RPS_NO_CPU 0xffff 705 706 extern u32 rps_cpu_mask; 707 extern struct rps_sock_flow_table __rcu *rps_sock_flow_table; 708 709 static inline void rps_record_sock_flow(struct rps_sock_flow_table *table, 710 u32 hash) 711 { 712 if (table && hash) { 713 unsigned int index = hash & table->mask; 714 u32 val = hash & ~rps_cpu_mask; 715 716 /* We only give a hint, preemption can change CPU under us */ 717 val |= raw_smp_processor_id(); 718 719 if (table->ents[index] != val) 720 table->ents[index] = val; 721 } 722 } 723 724 #ifdef CONFIG_RFS_ACCEL 725 bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index, u32 flow_id, 726 u16 filter_id); 727 #endif 728 #endif /* CONFIG_RPS */ 729 730 /* This structure contains an instance of an RX queue. */ 731 struct netdev_rx_queue { 732 struct xdp_rxq_info xdp_rxq; 733 #ifdef CONFIG_RPS 734 struct rps_map __rcu *rps_map; 735 struct rps_dev_flow_table __rcu *rps_flow_table; 736 #endif 737 struct kobject kobj; 738 struct net_device *dev; 739 netdevice_tracker dev_tracker; 740 741 #ifdef CONFIG_XDP_SOCKETS 742 struct xsk_buff_pool *pool; 743 #endif 744 } ____cacheline_aligned_in_smp; 745 746 /* 747 * RX queue sysfs structures and functions. 748 */ 749 struct rx_queue_attribute { 750 struct attribute attr; 751 ssize_t (*show)(struct netdev_rx_queue *queue, char *buf); 752 ssize_t (*store)(struct netdev_rx_queue *queue, 753 const char *buf, size_t len); 754 }; 755 756 /* XPS map type and offset of the xps map within net_device->xps_maps[]. */ 757 enum xps_map_type { 758 XPS_CPUS = 0, 759 XPS_RXQS, 760 XPS_MAPS_MAX, 761 }; 762 763 #ifdef CONFIG_XPS 764 /* 765 * This structure holds an XPS map which can be of variable length. The 766 * map is an array of queues. 767 */ 768 struct xps_map { 769 unsigned int len; 770 unsigned int alloc_len; 771 struct rcu_head rcu; 772 u16 queues[]; 773 }; 774 #define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16))) 775 #define XPS_MIN_MAP_ALLOC ((L1_CACHE_ALIGN(offsetof(struct xps_map, queues[1])) \ 776 - sizeof(struct xps_map)) / sizeof(u16)) 777 778 /* 779 * This structure holds all XPS maps for device. Maps are indexed by CPU. 780 * 781 * We keep track of the number of cpus/rxqs used when the struct is allocated, 782 * in nr_ids. This will help not accessing out-of-bound memory. 783 * 784 * We keep track of the number of traffic classes used when the struct is 785 * allocated, in num_tc. This will be used to navigate the maps, to ensure we're 786 * not crossing its upper bound, as the original dev->num_tc can be updated in 787 * the meantime. 788 */ 789 struct xps_dev_maps { 790 struct rcu_head rcu; 791 unsigned int nr_ids; 792 s16 num_tc; 793 struct xps_map __rcu *attr_map[]; /* Either CPUs map or RXQs map */ 794 }; 795 796 #define XPS_CPU_DEV_MAPS_SIZE(_tcs) (sizeof(struct xps_dev_maps) + \ 797 (nr_cpu_ids * (_tcs) * sizeof(struct xps_map *))) 798 799 #define XPS_RXQ_DEV_MAPS_SIZE(_tcs, _rxqs) (sizeof(struct xps_dev_maps) +\ 800 (_rxqs * (_tcs) * sizeof(struct xps_map *))) 801 802 #endif /* CONFIG_XPS */ 803 804 #define TC_MAX_QUEUE 16 805 #define TC_BITMASK 15 806 /* HW offloaded queuing disciplines txq count and offset maps */ 807 struct netdev_tc_txq { 808 u16 count; 809 u16 offset; 810 }; 811 812 #if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE) 813 /* 814 * This structure is to hold information about the device 815 * configured to run FCoE protocol stack. 816 */ 817 struct netdev_fcoe_hbainfo { 818 char manufacturer[64]; 819 char serial_number[64]; 820 char hardware_version[64]; 821 char driver_version[64]; 822 char optionrom_version[64]; 823 char firmware_version[64]; 824 char model[256]; 825 char model_description[256]; 826 }; 827 #endif 828 829 #define MAX_PHYS_ITEM_ID_LEN 32 830 831 /* This structure holds a unique identifier to identify some 832 * physical item (port for example) used by a netdevice. 833 */ 834 struct netdev_phys_item_id { 835 unsigned char id[MAX_PHYS_ITEM_ID_LEN]; 836 unsigned char id_len; 837 }; 838 839 static inline bool netdev_phys_item_id_same(struct netdev_phys_item_id *a, 840 struct netdev_phys_item_id *b) 841 { 842 return a->id_len == b->id_len && 843 memcmp(a->id, b->id, a->id_len) == 0; 844 } 845 846 typedef u16 (*select_queue_fallback_t)(struct net_device *dev, 847 struct sk_buff *skb, 848 struct net_device *sb_dev); 849 850 enum net_device_path_type { 851 DEV_PATH_ETHERNET = 0, 852 DEV_PATH_VLAN, 853 DEV_PATH_BRIDGE, 854 DEV_PATH_PPPOE, 855 DEV_PATH_DSA, 856 }; 857 858 struct net_device_path { 859 enum net_device_path_type type; 860 const struct net_device *dev; 861 union { 862 struct { 863 u16 id; 864 __be16 proto; 865 u8 h_dest[ETH_ALEN]; 866 } encap; 867 struct { 868 enum { 869 DEV_PATH_BR_VLAN_KEEP, 870 DEV_PATH_BR_VLAN_TAG, 871 DEV_PATH_BR_VLAN_UNTAG, 872 DEV_PATH_BR_VLAN_UNTAG_HW, 873 } vlan_mode; 874 u16 vlan_id; 875 __be16 vlan_proto; 876 } bridge; 877 struct { 878 int port; 879 u16 proto; 880 } dsa; 881 }; 882 }; 883 884 #define NET_DEVICE_PATH_STACK_MAX 5 885 #define NET_DEVICE_PATH_VLAN_MAX 2 886 887 struct net_device_path_stack { 888 int num_paths; 889 struct net_device_path path[NET_DEVICE_PATH_STACK_MAX]; 890 }; 891 892 struct net_device_path_ctx { 893 const struct net_device *dev; 894 const u8 *daddr; 895 896 int num_vlans; 897 struct { 898 u16 id; 899 __be16 proto; 900 } vlan[NET_DEVICE_PATH_VLAN_MAX]; 901 }; 902 903 enum tc_setup_type { 904 TC_SETUP_QDISC_MQPRIO, 905 TC_SETUP_CLSU32, 906 TC_SETUP_CLSFLOWER, 907 TC_SETUP_CLSMATCHALL, 908 TC_SETUP_CLSBPF, 909 TC_SETUP_BLOCK, 910 TC_SETUP_QDISC_CBS, 911 TC_SETUP_QDISC_RED, 912 TC_SETUP_QDISC_PRIO, 913 TC_SETUP_QDISC_MQ, 914 TC_SETUP_QDISC_ETF, 915 TC_SETUP_ROOT_QDISC, 916 TC_SETUP_QDISC_GRED, 917 TC_SETUP_QDISC_TAPRIO, 918 TC_SETUP_FT, 919 TC_SETUP_QDISC_ETS, 920 TC_SETUP_QDISC_TBF, 921 TC_SETUP_QDISC_FIFO, 922 TC_SETUP_QDISC_HTB, 923 }; 924 925 /* These structures hold the attributes of bpf state that are being passed 926 * to the netdevice through the bpf op. 927 */ 928 enum bpf_netdev_command { 929 /* Set or clear a bpf program used in the earliest stages of packet 930 * rx. The prog will have been loaded as BPF_PROG_TYPE_XDP. The callee 931 * is responsible for calling bpf_prog_put on any old progs that are 932 * stored. In case of error, the callee need not release the new prog 933 * reference, but on success it takes ownership and must bpf_prog_put 934 * when it is no longer used. 935 */ 936 XDP_SETUP_PROG, 937 XDP_SETUP_PROG_HW, 938 /* BPF program for offload callbacks, invoked at program load time. */ 939 BPF_OFFLOAD_MAP_ALLOC, 940 BPF_OFFLOAD_MAP_FREE, 941 XDP_SETUP_XSK_POOL, 942 }; 943 944 struct bpf_prog_offload_ops; 945 struct netlink_ext_ack; 946 struct xdp_umem; 947 struct xdp_dev_bulk_queue; 948 struct bpf_xdp_link; 949 950 enum bpf_xdp_mode { 951 XDP_MODE_SKB = 0, 952 XDP_MODE_DRV = 1, 953 XDP_MODE_HW = 2, 954 __MAX_XDP_MODE 955 }; 956 957 struct bpf_xdp_entity { 958 struct bpf_prog *prog; 959 struct bpf_xdp_link *link; 960 }; 961 962 struct netdev_bpf { 963 enum bpf_netdev_command command; 964 union { 965 /* XDP_SETUP_PROG */ 966 struct { 967 u32 flags; 968 struct bpf_prog *prog; 969 struct netlink_ext_ack *extack; 970 }; 971 /* BPF_OFFLOAD_MAP_ALLOC, BPF_OFFLOAD_MAP_FREE */ 972 struct { 973 struct bpf_offloaded_map *offmap; 974 }; 975 /* XDP_SETUP_XSK_POOL */ 976 struct { 977 struct xsk_buff_pool *pool; 978 u16 queue_id; 979 } xsk; 980 }; 981 }; 982 983 /* Flags for ndo_xsk_wakeup. */ 984 #define XDP_WAKEUP_RX (1 << 0) 985 #define XDP_WAKEUP_TX (1 << 1) 986 987 #ifdef CONFIG_XFRM_OFFLOAD 988 struct xfrmdev_ops { 989 int (*xdo_dev_state_add) (struct xfrm_state *x); 990 void (*xdo_dev_state_delete) (struct xfrm_state *x); 991 void (*xdo_dev_state_free) (struct xfrm_state *x); 992 bool (*xdo_dev_offload_ok) (struct sk_buff *skb, 993 struct xfrm_state *x); 994 void (*xdo_dev_state_advance_esn) (struct xfrm_state *x); 995 }; 996 #endif 997 998 struct dev_ifalias { 999 struct rcu_head rcuhead; 1000 char ifalias[]; 1001 }; 1002 1003 struct devlink; 1004 struct tlsdev_ops; 1005 1006 struct netdev_name_node { 1007 struct hlist_node hlist; 1008 struct list_head list; 1009 struct net_device *dev; 1010 const char *name; 1011 }; 1012 1013 int netdev_name_node_alt_create(struct net_device *dev, const char *name); 1014 int netdev_name_node_alt_destroy(struct net_device *dev, const char *name); 1015 1016 struct netdev_net_notifier { 1017 struct list_head list; 1018 struct notifier_block *nb; 1019 }; 1020 1021 /* 1022 * This structure defines the management hooks for network devices. 1023 * The following hooks can be defined; unless noted otherwise, they are 1024 * optional and can be filled with a null pointer. 1025 * 1026 * int (*ndo_init)(struct net_device *dev); 1027 * This function is called once when a network device is registered. 1028 * The network device can use this for any late stage initialization 1029 * or semantic validation. It can fail with an error code which will 1030 * be propagated back to register_netdev. 1031 * 1032 * void (*ndo_uninit)(struct net_device *dev); 1033 * This function is called when device is unregistered or when registration 1034 * fails. It is not called if init fails. 1035 * 1036 * int (*ndo_open)(struct net_device *dev); 1037 * This function is called when a network device transitions to the up 1038 * state. 1039 * 1040 * int (*ndo_stop)(struct net_device *dev); 1041 * This function is called when a network device transitions to the down 1042 * state. 1043 * 1044 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb, 1045 * struct net_device *dev); 1046 * Called when a packet needs to be transmitted. 1047 * Returns NETDEV_TX_OK. Can return NETDEV_TX_BUSY, but you should stop 1048 * the queue before that can happen; it's for obsolete devices and weird 1049 * corner cases, but the stack really does a non-trivial amount 1050 * of useless work if you return NETDEV_TX_BUSY. 1051 * Required; cannot be NULL. 1052 * 1053 * netdev_features_t (*ndo_features_check)(struct sk_buff *skb, 1054 * struct net_device *dev 1055 * netdev_features_t features); 1056 * Called by core transmit path to determine if device is capable of 1057 * performing offload operations on a given packet. This is to give 1058 * the device an opportunity to implement any restrictions that cannot 1059 * be otherwise expressed by feature flags. The check is called with 1060 * the set of features that the stack has calculated and it returns 1061 * those the driver believes to be appropriate. 1062 * 1063 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb, 1064 * struct net_device *sb_dev); 1065 * Called to decide which queue to use when device supports multiple 1066 * transmit queues. 1067 * 1068 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags); 1069 * This function is called to allow device receiver to make 1070 * changes to configuration when multicast or promiscuous is enabled. 1071 * 1072 * void (*ndo_set_rx_mode)(struct net_device *dev); 1073 * This function is called device changes address list filtering. 1074 * If driver handles unicast address filtering, it should set 1075 * IFF_UNICAST_FLT in its priv_flags. 1076 * 1077 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr); 1078 * This function is called when the Media Access Control address 1079 * needs to be changed. If this interface is not defined, the 1080 * MAC address can not be changed. 1081 * 1082 * int (*ndo_validate_addr)(struct net_device *dev); 1083 * Test if Media Access Control address is valid for the device. 1084 * 1085 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd); 1086 * Old-style ioctl entry point. This is used internally by the 1087 * appletalk and ieee802154 subsystems but is no longer called by 1088 * the device ioctl handler. 1089 * 1090 * int (*ndo_siocbond)(struct net_device *dev, struct ifreq *ifr, int cmd); 1091 * Used by the bonding driver for its device specific ioctls: 1092 * SIOCBONDENSLAVE, SIOCBONDRELEASE, SIOCBONDSETHWADDR, SIOCBONDCHANGEACTIVE, 1093 * SIOCBONDSLAVEINFOQUERY, and SIOCBONDINFOQUERY 1094 * 1095 * * int (*ndo_eth_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd); 1096 * Called for ethernet specific ioctls: SIOCGMIIPHY, SIOCGMIIREG, 1097 * SIOCSMIIREG, SIOCSHWTSTAMP and SIOCGHWTSTAMP. 1098 * 1099 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map); 1100 * Used to set network devices bus interface parameters. This interface 1101 * is retained for legacy reasons; new devices should use the bus 1102 * interface (PCI) for low level management. 1103 * 1104 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu); 1105 * Called when a user wants to change the Maximum Transfer Unit 1106 * of a device. 1107 * 1108 * void (*ndo_tx_timeout)(struct net_device *dev, unsigned int txqueue); 1109 * Callback used when the transmitter has not made any progress 1110 * for dev->watchdog ticks. 1111 * 1112 * void (*ndo_get_stats64)(struct net_device *dev, 1113 * struct rtnl_link_stats64 *storage); 1114 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev); 1115 * Called when a user wants to get the network device usage 1116 * statistics. Drivers must do one of the following: 1117 * 1. Define @ndo_get_stats64 to fill in a zero-initialised 1118 * rtnl_link_stats64 structure passed by the caller. 1119 * 2. Define @ndo_get_stats to update a net_device_stats structure 1120 * (which should normally be dev->stats) and return a pointer to 1121 * it. The structure may be changed asynchronously only if each 1122 * field is written atomically. 1123 * 3. Update dev->stats asynchronously and atomically, and define 1124 * neither operation. 1125 * 1126 * bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id) 1127 * Return true if this device supports offload stats of this attr_id. 1128 * 1129 * int (*ndo_get_offload_stats)(int attr_id, const struct net_device *dev, 1130 * void *attr_data) 1131 * Get statistics for offload operations by attr_id. Write it into the 1132 * attr_data pointer. 1133 * 1134 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16 vid); 1135 * If device supports VLAN filtering this function is called when a 1136 * VLAN id is registered. 1137 * 1138 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, __be16 proto, u16 vid); 1139 * If device supports VLAN filtering this function is called when a 1140 * VLAN id is unregistered. 1141 * 1142 * void (*ndo_poll_controller)(struct net_device *dev); 1143 * 1144 * SR-IOV management functions. 1145 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac); 1146 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan, 1147 * u8 qos, __be16 proto); 1148 * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate, 1149 * int max_tx_rate); 1150 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting); 1151 * int (*ndo_set_vf_trust)(struct net_device *dev, int vf, bool setting); 1152 * int (*ndo_get_vf_config)(struct net_device *dev, 1153 * int vf, struct ifla_vf_info *ivf); 1154 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state); 1155 * int (*ndo_set_vf_port)(struct net_device *dev, int vf, 1156 * struct nlattr *port[]); 1157 * 1158 * Enable or disable the VF ability to query its RSS Redirection Table and 1159 * Hash Key. This is needed since on some devices VF share this information 1160 * with PF and querying it may introduce a theoretical security risk. 1161 * int (*ndo_set_vf_rss_query_en)(struct net_device *dev, int vf, bool setting); 1162 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb); 1163 * int (*ndo_setup_tc)(struct net_device *dev, enum tc_setup_type type, 1164 * void *type_data); 1165 * Called to setup any 'tc' scheduler, classifier or action on @dev. 1166 * This is always called from the stack with the rtnl lock held and netif 1167 * tx queues stopped. This allows the netdevice to perform queue 1168 * management safely. 1169 * 1170 * Fiber Channel over Ethernet (FCoE) offload functions. 1171 * int (*ndo_fcoe_enable)(struct net_device *dev); 1172 * Called when the FCoE protocol stack wants to start using LLD for FCoE 1173 * so the underlying device can perform whatever needed configuration or 1174 * initialization to support acceleration of FCoE traffic. 1175 * 1176 * int (*ndo_fcoe_disable)(struct net_device *dev); 1177 * Called when the FCoE protocol stack wants to stop using LLD for FCoE 1178 * so the underlying device can perform whatever needed clean-ups to 1179 * stop supporting acceleration of FCoE traffic. 1180 * 1181 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid, 1182 * struct scatterlist *sgl, unsigned int sgc); 1183 * Called when the FCoE Initiator wants to initialize an I/O that 1184 * is a possible candidate for Direct Data Placement (DDP). The LLD can 1185 * perform necessary setup and returns 1 to indicate the device is set up 1186 * successfully to perform DDP on this I/O, otherwise this returns 0. 1187 * 1188 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid); 1189 * Called when the FCoE Initiator/Target is done with the DDPed I/O as 1190 * indicated by the FC exchange id 'xid', so the underlying device can 1191 * clean up and reuse resources for later DDP requests. 1192 * 1193 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid, 1194 * struct scatterlist *sgl, unsigned int sgc); 1195 * Called when the FCoE Target wants to initialize an I/O that 1196 * is a possible candidate for Direct Data Placement (DDP). The LLD can 1197 * perform necessary setup and returns 1 to indicate the device is set up 1198 * successfully to perform DDP on this I/O, otherwise this returns 0. 1199 * 1200 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev, 1201 * struct netdev_fcoe_hbainfo *hbainfo); 1202 * Called when the FCoE Protocol stack wants information on the underlying 1203 * device. This information is utilized by the FCoE protocol stack to 1204 * register attributes with Fiber Channel management service as per the 1205 * FC-GS Fabric Device Management Information(FDMI) specification. 1206 * 1207 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type); 1208 * Called when the underlying device wants to override default World Wide 1209 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own 1210 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE 1211 * protocol stack to use. 1212 * 1213 * RFS acceleration. 1214 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb, 1215 * u16 rxq_index, u32 flow_id); 1216 * Set hardware filter for RFS. rxq_index is the target queue index; 1217 * flow_id is a flow ID to be passed to rps_may_expire_flow() later. 1218 * Return the filter ID on success, or a negative error code. 1219 * 1220 * Slave management functions (for bridge, bonding, etc). 1221 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev); 1222 * Called to make another netdev an underling. 1223 * 1224 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev); 1225 * Called to release previously enslaved netdev. 1226 * 1227 * struct net_device *(*ndo_get_xmit_slave)(struct net_device *dev, 1228 * struct sk_buff *skb, 1229 * bool all_slaves); 1230 * Get the xmit slave of master device. If all_slaves is true, function 1231 * assume all the slaves can transmit. 1232 * 1233 * Feature/offload setting functions. 1234 * netdev_features_t (*ndo_fix_features)(struct net_device *dev, 1235 * netdev_features_t features); 1236 * Adjusts the requested feature flags according to device-specific 1237 * constraints, and returns the resulting flags. Must not modify 1238 * the device state. 1239 * 1240 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features); 1241 * Called to update device configuration to new features. Passed 1242 * feature set might be less than what was returned by ndo_fix_features()). 1243 * Must return >0 or -errno if it changed dev->features itself. 1244 * 1245 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[], 1246 * struct net_device *dev, 1247 * const unsigned char *addr, u16 vid, u16 flags, 1248 * struct netlink_ext_ack *extack); 1249 * Adds an FDB entry to dev for addr. 1250 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[], 1251 * struct net_device *dev, 1252 * const unsigned char *addr, u16 vid) 1253 * Deletes the FDB entry from dev coresponding to addr. 1254 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb, 1255 * struct net_device *dev, struct net_device *filter_dev, 1256 * int *idx) 1257 * Used to add FDB entries to dump requests. Implementers should add 1258 * entries to skb and update idx with the number of entries. 1259 * 1260 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh, 1261 * u16 flags, struct netlink_ext_ack *extack) 1262 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq, 1263 * struct net_device *dev, u32 filter_mask, 1264 * int nlflags) 1265 * int (*ndo_bridge_dellink)(struct net_device *dev, struct nlmsghdr *nlh, 1266 * u16 flags); 1267 * 1268 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier); 1269 * Called to change device carrier. Soft-devices (like dummy, team, etc) 1270 * which do not represent real hardware may define this to allow their 1271 * userspace components to manage their virtual carrier state. Devices 1272 * that determine carrier state from physical hardware properties (eg 1273 * network cables) or protocol-dependent mechanisms (eg 1274 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function. 1275 * 1276 * int (*ndo_get_phys_port_id)(struct net_device *dev, 1277 * struct netdev_phys_item_id *ppid); 1278 * Called to get ID of physical port of this device. If driver does 1279 * not implement this, it is assumed that the hw is not able to have 1280 * multiple net devices on single physical port. 1281 * 1282 * int (*ndo_get_port_parent_id)(struct net_device *dev, 1283 * struct netdev_phys_item_id *ppid) 1284 * Called to get the parent ID of the physical port of this device. 1285 * 1286 * void* (*ndo_dfwd_add_station)(struct net_device *pdev, 1287 * struct net_device *dev) 1288 * Called by upper layer devices to accelerate switching or other 1289 * station functionality into hardware. 'pdev is the lowerdev 1290 * to use for the offload and 'dev' is the net device that will 1291 * back the offload. Returns a pointer to the private structure 1292 * the upper layer will maintain. 1293 * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv) 1294 * Called by upper layer device to delete the station created 1295 * by 'ndo_dfwd_add_station'. 'pdev' is the net device backing 1296 * the station and priv is the structure returned by the add 1297 * operation. 1298 * int (*ndo_set_tx_maxrate)(struct net_device *dev, 1299 * int queue_index, u32 maxrate); 1300 * Called when a user wants to set a max-rate limitation of specific 1301 * TX queue. 1302 * int (*ndo_get_iflink)(const struct net_device *dev); 1303 * Called to get the iflink value of this device. 1304 * int (*ndo_fill_metadata_dst)(struct net_device *dev, struct sk_buff *skb); 1305 * This function is used to get egress tunnel information for given skb. 1306 * This is useful for retrieving outer tunnel header parameters while 1307 * sampling packet. 1308 * void (*ndo_set_rx_headroom)(struct net_device *dev, int needed_headroom); 1309 * This function is used to specify the headroom that the skb must 1310 * consider when allocation skb during packet reception. Setting 1311 * appropriate rx headroom value allows avoiding skb head copy on 1312 * forward. Setting a negative value resets the rx headroom to the 1313 * default value. 1314 * int (*ndo_bpf)(struct net_device *dev, struct netdev_bpf *bpf); 1315 * This function is used to set or query state related to XDP on the 1316 * netdevice and manage BPF offload. See definition of 1317 * enum bpf_netdev_command for details. 1318 * int (*ndo_xdp_xmit)(struct net_device *dev, int n, struct xdp_frame **xdp, 1319 * u32 flags); 1320 * This function is used to submit @n XDP packets for transmit on a 1321 * netdevice. Returns number of frames successfully transmitted, frames 1322 * that got dropped are freed/returned via xdp_return_frame(). 1323 * Returns negative number, means general error invoking ndo, meaning 1324 * no frames were xmit'ed and core-caller will free all frames. 1325 * struct net_device *(*ndo_xdp_get_xmit_slave)(struct net_device *dev, 1326 * struct xdp_buff *xdp); 1327 * Get the xmit slave of master device based on the xdp_buff. 1328 * int (*ndo_xsk_wakeup)(struct net_device *dev, u32 queue_id, u32 flags); 1329 * This function is used to wake up the softirq, ksoftirqd or kthread 1330 * responsible for sending and/or receiving packets on a specific 1331 * queue id bound to an AF_XDP socket. The flags field specifies if 1332 * only RX, only Tx, or both should be woken up using the flags 1333 * XDP_WAKEUP_RX and XDP_WAKEUP_TX. 1334 * struct devlink_port *(*ndo_get_devlink_port)(struct net_device *dev); 1335 * Get devlink port instance associated with a given netdev. 1336 * Called with a reference on the netdevice and devlink locks only, 1337 * rtnl_lock is not held. 1338 * int (*ndo_tunnel_ctl)(struct net_device *dev, struct ip_tunnel_parm *p, 1339 * int cmd); 1340 * Add, change, delete or get information on an IPv4 tunnel. 1341 * struct net_device *(*ndo_get_peer_dev)(struct net_device *dev); 1342 * If a device is paired with a peer device, return the peer instance. 1343 * The caller must be under RCU read context. 1344 * int (*ndo_fill_forward_path)(struct net_device_path_ctx *ctx, struct net_device_path *path); 1345 * Get the forwarding path to reach the real device from the HW destination address 1346 */ 1347 struct net_device_ops { 1348 int (*ndo_init)(struct net_device *dev); 1349 void (*ndo_uninit)(struct net_device *dev); 1350 int (*ndo_open)(struct net_device *dev); 1351 int (*ndo_stop)(struct net_device *dev); 1352 netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb, 1353 struct net_device *dev); 1354 netdev_features_t (*ndo_features_check)(struct sk_buff *skb, 1355 struct net_device *dev, 1356 netdev_features_t features); 1357 u16 (*ndo_select_queue)(struct net_device *dev, 1358 struct sk_buff *skb, 1359 struct net_device *sb_dev); 1360 void (*ndo_change_rx_flags)(struct net_device *dev, 1361 int flags); 1362 void (*ndo_set_rx_mode)(struct net_device *dev); 1363 int (*ndo_set_mac_address)(struct net_device *dev, 1364 void *addr); 1365 int (*ndo_validate_addr)(struct net_device *dev); 1366 int (*ndo_do_ioctl)(struct net_device *dev, 1367 struct ifreq *ifr, int cmd); 1368 int (*ndo_eth_ioctl)(struct net_device *dev, 1369 struct ifreq *ifr, int cmd); 1370 int (*ndo_siocbond)(struct net_device *dev, 1371 struct ifreq *ifr, int cmd); 1372 int (*ndo_siocwandev)(struct net_device *dev, 1373 struct if_settings *ifs); 1374 int (*ndo_siocdevprivate)(struct net_device *dev, 1375 struct ifreq *ifr, 1376 void __user *data, int cmd); 1377 int (*ndo_set_config)(struct net_device *dev, 1378 struct ifmap *map); 1379 int (*ndo_change_mtu)(struct net_device *dev, 1380 int new_mtu); 1381 int (*ndo_neigh_setup)(struct net_device *dev, 1382 struct neigh_parms *); 1383 void (*ndo_tx_timeout) (struct net_device *dev, 1384 unsigned int txqueue); 1385 1386 void (*ndo_get_stats64)(struct net_device *dev, 1387 struct rtnl_link_stats64 *storage); 1388 bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id); 1389 int (*ndo_get_offload_stats)(int attr_id, 1390 const struct net_device *dev, 1391 void *attr_data); 1392 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev); 1393 1394 int (*ndo_vlan_rx_add_vid)(struct net_device *dev, 1395 __be16 proto, u16 vid); 1396 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, 1397 __be16 proto, u16 vid); 1398 #ifdef CONFIG_NET_POLL_CONTROLLER 1399 void (*ndo_poll_controller)(struct net_device *dev); 1400 int (*ndo_netpoll_setup)(struct net_device *dev, 1401 struct netpoll_info *info); 1402 void (*ndo_netpoll_cleanup)(struct net_device *dev); 1403 #endif 1404 int (*ndo_set_vf_mac)(struct net_device *dev, 1405 int queue, u8 *mac); 1406 int (*ndo_set_vf_vlan)(struct net_device *dev, 1407 int queue, u16 vlan, 1408 u8 qos, __be16 proto); 1409 int (*ndo_set_vf_rate)(struct net_device *dev, 1410 int vf, int min_tx_rate, 1411 int max_tx_rate); 1412 int (*ndo_set_vf_spoofchk)(struct net_device *dev, 1413 int vf, bool setting); 1414 int (*ndo_set_vf_trust)(struct net_device *dev, 1415 int vf, bool setting); 1416 int (*ndo_get_vf_config)(struct net_device *dev, 1417 int vf, 1418 struct ifla_vf_info *ivf); 1419 int (*ndo_set_vf_link_state)(struct net_device *dev, 1420 int vf, int link_state); 1421 int (*ndo_get_vf_stats)(struct net_device *dev, 1422 int vf, 1423 struct ifla_vf_stats 1424 *vf_stats); 1425 int (*ndo_set_vf_port)(struct net_device *dev, 1426 int vf, 1427 struct nlattr *port[]); 1428 int (*ndo_get_vf_port)(struct net_device *dev, 1429 int vf, struct sk_buff *skb); 1430 int (*ndo_get_vf_guid)(struct net_device *dev, 1431 int vf, 1432 struct ifla_vf_guid *node_guid, 1433 struct ifla_vf_guid *port_guid); 1434 int (*ndo_set_vf_guid)(struct net_device *dev, 1435 int vf, u64 guid, 1436 int guid_type); 1437 int (*ndo_set_vf_rss_query_en)( 1438 struct net_device *dev, 1439 int vf, bool setting); 1440 int (*ndo_setup_tc)(struct net_device *dev, 1441 enum tc_setup_type type, 1442 void *type_data); 1443 #if IS_ENABLED(CONFIG_FCOE) 1444 int (*ndo_fcoe_enable)(struct net_device *dev); 1445 int (*ndo_fcoe_disable)(struct net_device *dev); 1446 int (*ndo_fcoe_ddp_setup)(struct net_device *dev, 1447 u16 xid, 1448 struct scatterlist *sgl, 1449 unsigned int sgc); 1450 int (*ndo_fcoe_ddp_done)(struct net_device *dev, 1451 u16 xid); 1452 int (*ndo_fcoe_ddp_target)(struct net_device *dev, 1453 u16 xid, 1454 struct scatterlist *sgl, 1455 unsigned int sgc); 1456 int (*ndo_fcoe_get_hbainfo)(struct net_device *dev, 1457 struct netdev_fcoe_hbainfo *hbainfo); 1458 #endif 1459 1460 #if IS_ENABLED(CONFIG_LIBFCOE) 1461 #define NETDEV_FCOE_WWNN 0 1462 #define NETDEV_FCOE_WWPN 1 1463 int (*ndo_fcoe_get_wwn)(struct net_device *dev, 1464 u64 *wwn, int type); 1465 #endif 1466 1467 #ifdef CONFIG_RFS_ACCEL 1468 int (*ndo_rx_flow_steer)(struct net_device *dev, 1469 const struct sk_buff *skb, 1470 u16 rxq_index, 1471 u32 flow_id); 1472 #endif 1473 int (*ndo_add_slave)(struct net_device *dev, 1474 struct net_device *slave_dev, 1475 struct netlink_ext_ack *extack); 1476 int (*ndo_del_slave)(struct net_device *dev, 1477 struct net_device *slave_dev); 1478 struct net_device* (*ndo_get_xmit_slave)(struct net_device *dev, 1479 struct sk_buff *skb, 1480 bool all_slaves); 1481 struct net_device* (*ndo_sk_get_lower_dev)(struct net_device *dev, 1482 struct sock *sk); 1483 netdev_features_t (*ndo_fix_features)(struct net_device *dev, 1484 netdev_features_t features); 1485 int (*ndo_set_features)(struct net_device *dev, 1486 netdev_features_t features); 1487 int (*ndo_neigh_construct)(struct net_device *dev, 1488 struct neighbour *n); 1489 void (*ndo_neigh_destroy)(struct net_device *dev, 1490 struct neighbour *n); 1491 1492 int (*ndo_fdb_add)(struct ndmsg *ndm, 1493 struct nlattr *tb[], 1494 struct net_device *dev, 1495 const unsigned char *addr, 1496 u16 vid, 1497 u16 flags, 1498 struct netlink_ext_ack *extack); 1499 int (*ndo_fdb_del)(struct ndmsg *ndm, 1500 struct nlattr *tb[], 1501 struct net_device *dev, 1502 const unsigned char *addr, 1503 u16 vid); 1504 int (*ndo_fdb_dump)(struct sk_buff *skb, 1505 struct netlink_callback *cb, 1506 struct net_device *dev, 1507 struct net_device *filter_dev, 1508 int *idx); 1509 int (*ndo_fdb_get)(struct sk_buff *skb, 1510 struct nlattr *tb[], 1511 struct net_device *dev, 1512 const unsigned char *addr, 1513 u16 vid, u32 portid, u32 seq, 1514 struct netlink_ext_ack *extack); 1515 int (*ndo_bridge_setlink)(struct net_device *dev, 1516 struct nlmsghdr *nlh, 1517 u16 flags, 1518 struct netlink_ext_ack *extack); 1519 int (*ndo_bridge_getlink)(struct sk_buff *skb, 1520 u32 pid, u32 seq, 1521 struct net_device *dev, 1522 u32 filter_mask, 1523 int nlflags); 1524 int (*ndo_bridge_dellink)(struct net_device *dev, 1525 struct nlmsghdr *nlh, 1526 u16 flags); 1527 int (*ndo_change_carrier)(struct net_device *dev, 1528 bool new_carrier); 1529 int (*ndo_get_phys_port_id)(struct net_device *dev, 1530 struct netdev_phys_item_id *ppid); 1531 int (*ndo_get_port_parent_id)(struct net_device *dev, 1532 struct netdev_phys_item_id *ppid); 1533 int (*ndo_get_phys_port_name)(struct net_device *dev, 1534 char *name, size_t len); 1535 void* (*ndo_dfwd_add_station)(struct net_device *pdev, 1536 struct net_device *dev); 1537 void (*ndo_dfwd_del_station)(struct net_device *pdev, 1538 void *priv); 1539 1540 int (*ndo_set_tx_maxrate)(struct net_device *dev, 1541 int queue_index, 1542 u32 maxrate); 1543 int (*ndo_get_iflink)(const struct net_device *dev); 1544 int (*ndo_fill_metadata_dst)(struct net_device *dev, 1545 struct sk_buff *skb); 1546 void (*ndo_set_rx_headroom)(struct net_device *dev, 1547 int needed_headroom); 1548 int (*ndo_bpf)(struct net_device *dev, 1549 struct netdev_bpf *bpf); 1550 int (*ndo_xdp_xmit)(struct net_device *dev, int n, 1551 struct xdp_frame **xdp, 1552 u32 flags); 1553 struct net_device * (*ndo_xdp_get_xmit_slave)(struct net_device *dev, 1554 struct xdp_buff *xdp); 1555 int (*ndo_xsk_wakeup)(struct net_device *dev, 1556 u32 queue_id, u32 flags); 1557 struct devlink_port * (*ndo_get_devlink_port)(struct net_device *dev); 1558 int (*ndo_tunnel_ctl)(struct net_device *dev, 1559 struct ip_tunnel_parm *p, int cmd); 1560 struct net_device * (*ndo_get_peer_dev)(struct net_device *dev); 1561 int (*ndo_fill_forward_path)(struct net_device_path_ctx *ctx, 1562 struct net_device_path *path); 1563 }; 1564 1565 /** 1566 * enum netdev_priv_flags - &struct net_device priv_flags 1567 * 1568 * These are the &struct net_device, they are only set internally 1569 * by drivers and used in the kernel. These flags are invisible to 1570 * userspace; this means that the order of these flags can change 1571 * during any kernel release. 1572 * 1573 * You should have a pretty good reason to be extending these flags. 1574 * 1575 * @IFF_802_1Q_VLAN: 802.1Q VLAN device 1576 * @IFF_EBRIDGE: Ethernet bridging device 1577 * @IFF_BONDING: bonding master or slave 1578 * @IFF_ISATAP: ISATAP interface (RFC4214) 1579 * @IFF_WAN_HDLC: WAN HDLC device 1580 * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to 1581 * release skb->dst 1582 * @IFF_DONT_BRIDGE: disallow bridging this ether dev 1583 * @IFF_DISABLE_NETPOLL: disable netpoll at run-time 1584 * @IFF_MACVLAN_PORT: device used as macvlan port 1585 * @IFF_BRIDGE_PORT: device used as bridge port 1586 * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port 1587 * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit 1588 * @IFF_UNICAST_FLT: Supports unicast filtering 1589 * @IFF_TEAM_PORT: device used as team port 1590 * @IFF_SUPP_NOFCS: device supports sending custom FCS 1591 * @IFF_LIVE_ADDR_CHANGE: device supports hardware address 1592 * change when it's running 1593 * @IFF_MACVLAN: Macvlan device 1594 * @IFF_XMIT_DST_RELEASE_PERM: IFF_XMIT_DST_RELEASE not taking into account 1595 * underlying stacked devices 1596 * @IFF_L3MDEV_MASTER: device is an L3 master device 1597 * @IFF_NO_QUEUE: device can run without qdisc attached 1598 * @IFF_OPENVSWITCH: device is a Open vSwitch master 1599 * @IFF_L3MDEV_SLAVE: device is enslaved to an L3 master device 1600 * @IFF_TEAM: device is a team device 1601 * @IFF_RXFH_CONFIGURED: device has had Rx Flow indirection table configured 1602 * @IFF_PHONY_HEADROOM: the headroom value is controlled by an external 1603 * entity (i.e. the master device for bridged veth) 1604 * @IFF_MACSEC: device is a MACsec device 1605 * @IFF_NO_RX_HANDLER: device doesn't support the rx_handler hook 1606 * @IFF_FAILOVER: device is a failover master device 1607 * @IFF_FAILOVER_SLAVE: device is lower dev of a failover master device 1608 * @IFF_L3MDEV_RX_HANDLER: only invoke the rx handler of L3 master device 1609 * @IFF_LIVE_RENAME_OK: rename is allowed while device is up and running 1610 * @IFF_TX_SKB_NO_LINEAR: device/driver is capable of xmitting frames with 1611 * skb_headlen(skb) == 0 (data starts from frag0) 1612 * @IFF_CHANGE_PROTO_DOWN: device supports setting carrier via IFLA_PROTO_DOWN 1613 */ 1614 enum netdev_priv_flags { 1615 IFF_802_1Q_VLAN = 1<<0, 1616 IFF_EBRIDGE = 1<<1, 1617 IFF_BONDING = 1<<2, 1618 IFF_ISATAP = 1<<3, 1619 IFF_WAN_HDLC = 1<<4, 1620 IFF_XMIT_DST_RELEASE = 1<<5, 1621 IFF_DONT_BRIDGE = 1<<6, 1622 IFF_DISABLE_NETPOLL = 1<<7, 1623 IFF_MACVLAN_PORT = 1<<8, 1624 IFF_BRIDGE_PORT = 1<<9, 1625 IFF_OVS_DATAPATH = 1<<10, 1626 IFF_TX_SKB_SHARING = 1<<11, 1627 IFF_UNICAST_FLT = 1<<12, 1628 IFF_TEAM_PORT = 1<<13, 1629 IFF_SUPP_NOFCS = 1<<14, 1630 IFF_LIVE_ADDR_CHANGE = 1<<15, 1631 IFF_MACVLAN = 1<<16, 1632 IFF_XMIT_DST_RELEASE_PERM = 1<<17, 1633 IFF_L3MDEV_MASTER = 1<<18, 1634 IFF_NO_QUEUE = 1<<19, 1635 IFF_OPENVSWITCH = 1<<20, 1636 IFF_L3MDEV_SLAVE = 1<<21, 1637 IFF_TEAM = 1<<22, 1638 IFF_RXFH_CONFIGURED = 1<<23, 1639 IFF_PHONY_HEADROOM = 1<<24, 1640 IFF_MACSEC = 1<<25, 1641 IFF_NO_RX_HANDLER = 1<<26, 1642 IFF_FAILOVER = 1<<27, 1643 IFF_FAILOVER_SLAVE = 1<<28, 1644 IFF_L3MDEV_RX_HANDLER = 1<<29, 1645 IFF_LIVE_RENAME_OK = 1<<30, 1646 IFF_TX_SKB_NO_LINEAR = 1<<31, 1647 IFF_CHANGE_PROTO_DOWN = BIT_ULL(32), 1648 }; 1649 1650 #define IFF_802_1Q_VLAN IFF_802_1Q_VLAN 1651 #define IFF_EBRIDGE IFF_EBRIDGE 1652 #define IFF_BONDING IFF_BONDING 1653 #define IFF_ISATAP IFF_ISATAP 1654 #define IFF_WAN_HDLC IFF_WAN_HDLC 1655 #define IFF_XMIT_DST_RELEASE IFF_XMIT_DST_RELEASE 1656 #define IFF_DONT_BRIDGE IFF_DONT_BRIDGE 1657 #define IFF_DISABLE_NETPOLL IFF_DISABLE_NETPOLL 1658 #define IFF_MACVLAN_PORT IFF_MACVLAN_PORT 1659 #define IFF_BRIDGE_PORT IFF_BRIDGE_PORT 1660 #define IFF_OVS_DATAPATH IFF_OVS_DATAPATH 1661 #define IFF_TX_SKB_SHARING IFF_TX_SKB_SHARING 1662 #define IFF_UNICAST_FLT IFF_UNICAST_FLT 1663 #define IFF_TEAM_PORT IFF_TEAM_PORT 1664 #define IFF_SUPP_NOFCS IFF_SUPP_NOFCS 1665 #define IFF_LIVE_ADDR_CHANGE IFF_LIVE_ADDR_CHANGE 1666 #define IFF_MACVLAN IFF_MACVLAN 1667 #define IFF_XMIT_DST_RELEASE_PERM IFF_XMIT_DST_RELEASE_PERM 1668 #define IFF_L3MDEV_MASTER IFF_L3MDEV_MASTER 1669 #define IFF_NO_QUEUE IFF_NO_QUEUE 1670 #define IFF_OPENVSWITCH IFF_OPENVSWITCH 1671 #define IFF_L3MDEV_SLAVE IFF_L3MDEV_SLAVE 1672 #define IFF_TEAM IFF_TEAM 1673 #define IFF_RXFH_CONFIGURED IFF_RXFH_CONFIGURED 1674 #define IFF_PHONY_HEADROOM IFF_PHONY_HEADROOM 1675 #define IFF_MACSEC IFF_MACSEC 1676 #define IFF_NO_RX_HANDLER IFF_NO_RX_HANDLER 1677 #define IFF_FAILOVER IFF_FAILOVER 1678 #define IFF_FAILOVER_SLAVE IFF_FAILOVER_SLAVE 1679 #define IFF_L3MDEV_RX_HANDLER IFF_L3MDEV_RX_HANDLER 1680 #define IFF_LIVE_RENAME_OK IFF_LIVE_RENAME_OK 1681 #define IFF_TX_SKB_NO_LINEAR IFF_TX_SKB_NO_LINEAR 1682 1683 /* Specifies the type of the struct net_device::ml_priv pointer */ 1684 enum netdev_ml_priv_type { 1685 ML_PRIV_NONE, 1686 ML_PRIV_CAN, 1687 }; 1688 1689 /** 1690 * struct net_device - The DEVICE structure. 1691 * 1692 * Actually, this whole structure is a big mistake. It mixes I/O 1693 * data with strictly "high-level" data, and it has to know about 1694 * almost every data structure used in the INET module. 1695 * 1696 * @name: This is the first field of the "visible" part of this structure 1697 * (i.e. as seen by users in the "Space.c" file). It is the name 1698 * of the interface. 1699 * 1700 * @name_node: Name hashlist node 1701 * @ifalias: SNMP alias 1702 * @mem_end: Shared memory end 1703 * @mem_start: Shared memory start 1704 * @base_addr: Device I/O address 1705 * @irq: Device IRQ number 1706 * 1707 * @state: Generic network queuing layer state, see netdev_state_t 1708 * @dev_list: The global list of network devices 1709 * @napi_list: List entry used for polling NAPI devices 1710 * @unreg_list: List entry when we are unregistering the 1711 * device; see the function unregister_netdev 1712 * @close_list: List entry used when we are closing the device 1713 * @ptype_all: Device-specific packet handlers for all protocols 1714 * @ptype_specific: Device-specific, protocol-specific packet handlers 1715 * 1716 * @adj_list: Directly linked devices, like slaves for bonding 1717 * @features: Currently active device features 1718 * @hw_features: User-changeable features 1719 * 1720 * @wanted_features: User-requested features 1721 * @vlan_features: Mask of features inheritable by VLAN devices 1722 * 1723 * @hw_enc_features: Mask of features inherited by encapsulating devices 1724 * This field indicates what encapsulation 1725 * offloads the hardware is capable of doing, 1726 * and drivers will need to set them appropriately. 1727 * 1728 * @mpls_features: Mask of features inheritable by MPLS 1729 * @gso_partial_features: value(s) from NETIF_F_GSO\* 1730 * 1731 * @ifindex: interface index 1732 * @group: The group the device belongs to 1733 * 1734 * @stats: Statistics struct, which was left as a legacy, use 1735 * rtnl_link_stats64 instead 1736 * 1737 * @rx_dropped: Dropped packets by core network, 1738 * do not use this in drivers 1739 * @tx_dropped: Dropped packets by core network, 1740 * do not use this in drivers 1741 * @rx_nohandler: nohandler dropped packets by core network on 1742 * inactive devices, do not use this in drivers 1743 * @carrier_up_count: Number of times the carrier has been up 1744 * @carrier_down_count: Number of times the carrier has been down 1745 * 1746 * @wireless_handlers: List of functions to handle Wireless Extensions, 1747 * instead of ioctl, 1748 * see <net/iw_handler.h> for details. 1749 * @wireless_data: Instance data managed by the core of wireless extensions 1750 * 1751 * @netdev_ops: Includes several pointers to callbacks, 1752 * if one wants to override the ndo_*() functions 1753 * @ethtool_ops: Management operations 1754 * @l3mdev_ops: Layer 3 master device operations 1755 * @ndisc_ops: Includes callbacks for different IPv6 neighbour 1756 * discovery handling. Necessary for e.g. 6LoWPAN. 1757 * @xfrmdev_ops: Transformation offload operations 1758 * @tlsdev_ops: Transport Layer Security offload operations 1759 * @header_ops: Includes callbacks for creating,parsing,caching,etc 1760 * of Layer 2 headers. 1761 * 1762 * @flags: Interface flags (a la BSD) 1763 * @priv_flags: Like 'flags' but invisible to userspace, 1764 * see if.h for the definitions 1765 * @gflags: Global flags ( kept as legacy ) 1766 * @padded: How much padding added by alloc_netdev() 1767 * @operstate: RFC2863 operstate 1768 * @link_mode: Mapping policy to operstate 1769 * @if_port: Selectable AUI, TP, ... 1770 * @dma: DMA channel 1771 * @mtu: Interface MTU value 1772 * @min_mtu: Interface Minimum MTU value 1773 * @max_mtu: Interface Maximum MTU value 1774 * @type: Interface hardware type 1775 * @hard_header_len: Maximum hardware header length. 1776 * @min_header_len: Minimum hardware header length 1777 * 1778 * @needed_headroom: Extra headroom the hardware may need, but not in all 1779 * cases can this be guaranteed 1780 * @needed_tailroom: Extra tailroom the hardware may need, but not in all 1781 * cases can this be guaranteed. Some cases also use 1782 * LL_MAX_HEADER instead to allocate the skb 1783 * 1784 * interface address info: 1785 * 1786 * @perm_addr: Permanent hw address 1787 * @addr_assign_type: Hw address assignment type 1788 * @addr_len: Hardware address length 1789 * @upper_level: Maximum depth level of upper devices. 1790 * @lower_level: Maximum depth level of lower devices. 1791 * @neigh_priv_len: Used in neigh_alloc() 1792 * @dev_id: Used to differentiate devices that share 1793 * the same link layer address 1794 * @dev_port: Used to differentiate devices that share 1795 * the same function 1796 * @addr_list_lock: XXX: need comments on this one 1797 * @name_assign_type: network interface name assignment type 1798 * @uc_promisc: Counter that indicates promiscuous mode 1799 * has been enabled due to the need to listen to 1800 * additional unicast addresses in a device that 1801 * does not implement ndo_set_rx_mode() 1802 * @uc: unicast mac addresses 1803 * @mc: multicast mac addresses 1804 * @dev_addrs: list of device hw addresses 1805 * @queues_kset: Group of all Kobjects in the Tx and RX queues 1806 * @promiscuity: Number of times the NIC is told to work in 1807 * promiscuous mode; if it becomes 0 the NIC will 1808 * exit promiscuous mode 1809 * @allmulti: Counter, enables or disables allmulticast mode 1810 * 1811 * @vlan_info: VLAN info 1812 * @dsa_ptr: dsa specific data 1813 * @tipc_ptr: TIPC specific data 1814 * @atalk_ptr: AppleTalk link 1815 * @ip_ptr: IPv4 specific data 1816 * @dn_ptr: DECnet specific data 1817 * @ip6_ptr: IPv6 specific data 1818 * @ax25_ptr: AX.25 specific data 1819 * @ieee80211_ptr: IEEE 802.11 specific data, assign before registering 1820 * @ieee802154_ptr: IEEE 802.15.4 low-rate Wireless Personal Area Network 1821 * device struct 1822 * @mpls_ptr: mpls_dev struct pointer 1823 * @mctp_ptr: MCTP specific data 1824 * 1825 * @dev_addr: Hw address (before bcast, 1826 * because most packets are unicast) 1827 * 1828 * @_rx: Array of RX queues 1829 * @num_rx_queues: Number of RX queues 1830 * allocated at register_netdev() time 1831 * @real_num_rx_queues: Number of RX queues currently active in device 1832 * @xdp_prog: XDP sockets filter program pointer 1833 * @gro_flush_timeout: timeout for GRO layer in NAPI 1834 * @napi_defer_hard_irqs: If not zero, provides a counter that would 1835 * allow to avoid NIC hard IRQ, on busy queues. 1836 * 1837 * @rx_handler: handler for received packets 1838 * @rx_handler_data: XXX: need comments on this one 1839 * @miniq_ingress: ingress/clsact qdisc specific data for 1840 * ingress processing 1841 * @ingress_queue: XXX: need comments on this one 1842 * @nf_hooks_ingress: netfilter hooks executed for ingress packets 1843 * @broadcast: hw bcast address 1844 * 1845 * @rx_cpu_rmap: CPU reverse-mapping for RX completion interrupts, 1846 * indexed by RX queue number. Assigned by driver. 1847 * This must only be set if the ndo_rx_flow_steer 1848 * operation is defined 1849 * @index_hlist: Device index hash chain 1850 * 1851 * @_tx: Array of TX queues 1852 * @num_tx_queues: Number of TX queues allocated at alloc_netdev_mq() time 1853 * @real_num_tx_queues: Number of TX queues currently active in device 1854 * @qdisc: Root qdisc from userspace point of view 1855 * @tx_queue_len: Max frames per queue allowed 1856 * @tx_global_lock: XXX: need comments on this one 1857 * @xdp_bulkq: XDP device bulk queue 1858 * @xps_maps: all CPUs/RXQs maps for XPS device 1859 * 1860 * @xps_maps: XXX: need comments on this one 1861 * @miniq_egress: clsact qdisc specific data for 1862 * egress processing 1863 * @nf_hooks_egress: netfilter hooks executed for egress packets 1864 * @qdisc_hash: qdisc hash table 1865 * @watchdog_timeo: Represents the timeout that is used by 1866 * the watchdog (see dev_watchdog()) 1867 * @watchdog_timer: List of timers 1868 * 1869 * @proto_down_reason: reason a netdev interface is held down 1870 * @pcpu_refcnt: Number of references to this device 1871 * @dev_refcnt: Number of references to this device 1872 * @refcnt_tracker: Tracker directory for tracked references to this device 1873 * @todo_list: Delayed register/unregister 1874 * @link_watch_list: XXX: need comments on this one 1875 * 1876 * @reg_state: Register/unregister state machine 1877 * @dismantle: Device is going to be freed 1878 * @rtnl_link_state: This enum represents the phases of creating 1879 * a new link 1880 * 1881 * @needs_free_netdev: Should unregister perform free_netdev? 1882 * @priv_destructor: Called from unregister 1883 * @npinfo: XXX: need comments on this one 1884 * @nd_net: Network namespace this network device is inside 1885 * 1886 * @ml_priv: Mid-layer private 1887 * @ml_priv_type: Mid-layer private type 1888 * @lstats: Loopback statistics 1889 * @tstats: Tunnel statistics 1890 * @dstats: Dummy statistics 1891 * @vstats: Virtual ethernet statistics 1892 * 1893 * @garp_port: GARP 1894 * @mrp_port: MRP 1895 * 1896 * @dev: Class/net/name entry 1897 * @sysfs_groups: Space for optional device, statistics and wireless 1898 * sysfs groups 1899 * 1900 * @sysfs_rx_queue_group: Space for optional per-rx queue attributes 1901 * @rtnl_link_ops: Rtnl_link_ops 1902 * 1903 * @gso_max_size: Maximum size of generic segmentation offload 1904 * @gso_max_segs: Maximum number of segments that can be passed to the 1905 * NIC for GSO 1906 * 1907 * @dcbnl_ops: Data Center Bridging netlink ops 1908 * @num_tc: Number of traffic classes in the net device 1909 * @tc_to_txq: XXX: need comments on this one 1910 * @prio_tc_map: XXX: need comments on this one 1911 * 1912 * @fcoe_ddp_xid: Max exchange id for FCoE LRO by ddp 1913 * 1914 * @priomap: XXX: need comments on this one 1915 * @phydev: Physical device may attach itself 1916 * for hardware timestamping 1917 * @sfp_bus: attached &struct sfp_bus structure. 1918 * 1919 * @qdisc_tx_busylock: lockdep class annotating Qdisc->busylock spinlock 1920 * 1921 * @proto_down: protocol port state information can be sent to the 1922 * switch driver and used to set the phys state of the 1923 * switch port. 1924 * 1925 * @wol_enabled: Wake-on-LAN is enabled 1926 * 1927 * @threaded: napi threaded mode is enabled 1928 * 1929 * @net_notifier_list: List of per-net netdev notifier block 1930 * that follow this device when it is moved 1931 * to another network namespace. 1932 * 1933 * @macsec_ops: MACsec offloading ops 1934 * 1935 * @udp_tunnel_nic_info: static structure describing the UDP tunnel 1936 * offload capabilities of the device 1937 * @udp_tunnel_nic: UDP tunnel offload state 1938 * @xdp_state: stores info on attached XDP BPF programs 1939 * 1940 * @nested_level: Used as as a parameter of spin_lock_nested() of 1941 * dev->addr_list_lock. 1942 * @unlink_list: As netif_addr_lock() can be called recursively, 1943 * keep a list of interfaces to be deleted. 1944 * 1945 * @dev_addr_shadow: Copy of @dev_addr to catch direct writes. 1946 * @linkwatch_dev_tracker: refcount tracker used by linkwatch. 1947 * @watchdog_dev_tracker: refcount tracker used by watchdog. 1948 * 1949 * FIXME: cleanup struct net_device such that network protocol info 1950 * moves out. 1951 */ 1952 1953 struct net_device { 1954 char name[IFNAMSIZ]; 1955 struct netdev_name_node *name_node; 1956 struct dev_ifalias __rcu *ifalias; 1957 /* 1958 * I/O specific fields 1959 * FIXME: Merge these and struct ifmap into one 1960 */ 1961 unsigned long mem_end; 1962 unsigned long mem_start; 1963 unsigned long base_addr; 1964 1965 /* 1966 * Some hardware also needs these fields (state,dev_list, 1967 * napi_list,unreg_list,close_list) but they are not 1968 * part of the usual set specified in Space.c. 1969 */ 1970 1971 unsigned long state; 1972 1973 struct list_head dev_list; 1974 struct list_head napi_list; 1975 struct list_head unreg_list; 1976 struct list_head close_list; 1977 struct list_head ptype_all; 1978 struct list_head ptype_specific; 1979 1980 struct { 1981 struct list_head upper; 1982 struct list_head lower; 1983 } adj_list; 1984 1985 /* Read-mostly cache-line for fast-path access */ 1986 unsigned int flags; 1987 unsigned long long priv_flags; 1988 const struct net_device_ops *netdev_ops; 1989 int ifindex; 1990 unsigned short gflags; 1991 unsigned short hard_header_len; 1992 1993 /* Note : dev->mtu is often read without holding a lock. 1994 * Writers usually hold RTNL. 1995 * It is recommended to use READ_ONCE() to annotate the reads, 1996 * and to use WRITE_ONCE() to annotate the writes. 1997 */ 1998 unsigned int mtu; 1999 unsigned short needed_headroom; 2000 unsigned short needed_tailroom; 2001 2002 netdev_features_t features; 2003 netdev_features_t hw_features; 2004 netdev_features_t wanted_features; 2005 netdev_features_t vlan_features; 2006 netdev_features_t hw_enc_features; 2007 netdev_features_t mpls_features; 2008 netdev_features_t gso_partial_features; 2009 2010 unsigned int min_mtu; 2011 unsigned int max_mtu; 2012 unsigned short type; 2013 unsigned char min_header_len; 2014 unsigned char name_assign_type; 2015 2016 int group; 2017 2018 struct net_device_stats stats; /* not used by modern drivers */ 2019 2020 atomic_long_t rx_dropped; 2021 atomic_long_t tx_dropped; 2022 atomic_long_t rx_nohandler; 2023 2024 /* Stats to monitor link on/off, flapping */ 2025 atomic_t carrier_up_count; 2026 atomic_t carrier_down_count; 2027 2028 #ifdef CONFIG_WIRELESS_EXT 2029 const struct iw_handler_def *wireless_handlers; 2030 struct iw_public_data *wireless_data; 2031 #endif 2032 const struct ethtool_ops *ethtool_ops; 2033 #ifdef CONFIG_NET_L3_MASTER_DEV 2034 const struct l3mdev_ops *l3mdev_ops; 2035 #endif 2036 #if IS_ENABLED(CONFIG_IPV6) 2037 const struct ndisc_ops *ndisc_ops; 2038 #endif 2039 2040 #ifdef CONFIG_XFRM_OFFLOAD 2041 const struct xfrmdev_ops *xfrmdev_ops; 2042 #endif 2043 2044 #if IS_ENABLED(CONFIG_TLS_DEVICE) 2045 const struct tlsdev_ops *tlsdev_ops; 2046 #endif 2047 2048 const struct header_ops *header_ops; 2049 2050 unsigned char operstate; 2051 unsigned char link_mode; 2052 2053 unsigned char if_port; 2054 unsigned char dma; 2055 2056 /* Interface address info. */ 2057 unsigned char perm_addr[MAX_ADDR_LEN]; 2058 unsigned char addr_assign_type; 2059 unsigned char addr_len; 2060 unsigned char upper_level; 2061 unsigned char lower_level; 2062 2063 unsigned short neigh_priv_len; 2064 unsigned short dev_id; 2065 unsigned short dev_port; 2066 unsigned short padded; 2067 2068 spinlock_t addr_list_lock; 2069 int irq; 2070 2071 struct netdev_hw_addr_list uc; 2072 struct netdev_hw_addr_list mc; 2073 struct netdev_hw_addr_list dev_addrs; 2074 2075 #ifdef CONFIG_SYSFS 2076 struct kset *queues_kset; 2077 #endif 2078 #ifdef CONFIG_LOCKDEP 2079 struct list_head unlink_list; 2080 #endif 2081 unsigned int promiscuity; 2082 unsigned int allmulti; 2083 bool uc_promisc; 2084 #ifdef CONFIG_LOCKDEP 2085 unsigned char nested_level; 2086 #endif 2087 2088 2089 /* Protocol-specific pointers */ 2090 2091 #if IS_ENABLED(CONFIG_VLAN_8021Q) 2092 struct vlan_info __rcu *vlan_info; 2093 #endif 2094 #if IS_ENABLED(CONFIG_NET_DSA) 2095 struct dsa_port *dsa_ptr; 2096 #endif 2097 #if IS_ENABLED(CONFIG_TIPC) 2098 struct tipc_bearer __rcu *tipc_ptr; 2099 #endif 2100 #if IS_ENABLED(CONFIG_IRDA) || IS_ENABLED(CONFIG_ATALK) 2101 void *atalk_ptr; 2102 #endif 2103 struct in_device __rcu *ip_ptr; 2104 #if IS_ENABLED(CONFIG_DECNET) 2105 struct dn_dev __rcu *dn_ptr; 2106 #endif 2107 struct inet6_dev __rcu *ip6_ptr; 2108 #if IS_ENABLED(CONFIG_AX25) 2109 void *ax25_ptr; 2110 #endif 2111 struct wireless_dev *ieee80211_ptr; 2112 struct wpan_dev *ieee802154_ptr; 2113 #if IS_ENABLED(CONFIG_MPLS_ROUTING) 2114 struct mpls_dev __rcu *mpls_ptr; 2115 #endif 2116 #if IS_ENABLED(CONFIG_MCTP) 2117 struct mctp_dev __rcu *mctp_ptr; 2118 #endif 2119 2120 /* 2121 * Cache lines mostly used on receive path (including eth_type_trans()) 2122 */ 2123 /* Interface address info used in eth_type_trans() */ 2124 const unsigned char *dev_addr; 2125 2126 struct netdev_rx_queue *_rx; 2127 unsigned int num_rx_queues; 2128 unsigned int real_num_rx_queues; 2129 2130 struct bpf_prog __rcu *xdp_prog; 2131 unsigned long gro_flush_timeout; 2132 int napi_defer_hard_irqs; 2133 rx_handler_func_t __rcu *rx_handler; 2134 void __rcu *rx_handler_data; 2135 2136 #ifdef CONFIG_NET_CLS_ACT 2137 struct mini_Qdisc __rcu *miniq_ingress; 2138 #endif 2139 struct netdev_queue __rcu *ingress_queue; 2140 #ifdef CONFIG_NETFILTER_INGRESS 2141 struct nf_hook_entries __rcu *nf_hooks_ingress; 2142 #endif 2143 2144 unsigned char broadcast[MAX_ADDR_LEN]; 2145 #ifdef CONFIG_RFS_ACCEL 2146 struct cpu_rmap *rx_cpu_rmap; 2147 #endif 2148 struct hlist_node index_hlist; 2149 2150 /* 2151 * Cache lines mostly used on transmit path 2152 */ 2153 struct netdev_queue *_tx ____cacheline_aligned_in_smp; 2154 unsigned int num_tx_queues; 2155 unsigned int real_num_tx_queues; 2156 struct Qdisc *qdisc; 2157 unsigned int tx_queue_len; 2158 spinlock_t tx_global_lock; 2159 2160 struct xdp_dev_bulk_queue __percpu *xdp_bulkq; 2161 2162 #ifdef CONFIG_XPS 2163 struct xps_dev_maps __rcu *xps_maps[XPS_MAPS_MAX]; 2164 #endif 2165 #ifdef CONFIG_NET_CLS_ACT 2166 struct mini_Qdisc __rcu *miniq_egress; 2167 #endif 2168 #ifdef CONFIG_NETFILTER_EGRESS 2169 struct nf_hook_entries __rcu *nf_hooks_egress; 2170 #endif 2171 2172 #ifdef CONFIG_NET_SCHED 2173 DECLARE_HASHTABLE (qdisc_hash, 4); 2174 #endif 2175 /* These may be needed for future network-power-down code. */ 2176 struct timer_list watchdog_timer; 2177 int watchdog_timeo; 2178 2179 u32 proto_down_reason; 2180 2181 struct list_head todo_list; 2182 2183 #ifdef CONFIG_PCPU_DEV_REFCNT 2184 int __percpu *pcpu_refcnt; 2185 #else 2186 refcount_t dev_refcnt; 2187 #endif 2188 struct ref_tracker_dir refcnt_tracker; 2189 2190 struct list_head link_watch_list; 2191 2192 enum { NETREG_UNINITIALIZED=0, 2193 NETREG_REGISTERED, /* completed register_netdevice */ 2194 NETREG_UNREGISTERING, /* called unregister_netdevice */ 2195 NETREG_UNREGISTERED, /* completed unregister todo */ 2196 NETREG_RELEASED, /* called free_netdev */ 2197 NETREG_DUMMY, /* dummy device for NAPI poll */ 2198 } reg_state:8; 2199 2200 bool dismantle; 2201 2202 enum { 2203 RTNL_LINK_INITIALIZED, 2204 RTNL_LINK_INITIALIZING, 2205 } rtnl_link_state:16; 2206 2207 bool needs_free_netdev; 2208 void (*priv_destructor)(struct net_device *dev); 2209 2210 #ifdef CONFIG_NETPOLL 2211 struct netpoll_info __rcu *npinfo; 2212 #endif 2213 2214 possible_net_t nd_net; 2215 2216 /* mid-layer private */ 2217 void *ml_priv; 2218 enum netdev_ml_priv_type ml_priv_type; 2219 2220 union { 2221 struct pcpu_lstats __percpu *lstats; 2222 struct pcpu_sw_netstats __percpu *tstats; 2223 struct pcpu_dstats __percpu *dstats; 2224 }; 2225 2226 #if IS_ENABLED(CONFIG_GARP) 2227 struct garp_port __rcu *garp_port; 2228 #endif 2229 #if IS_ENABLED(CONFIG_MRP) 2230 struct mrp_port __rcu *mrp_port; 2231 #endif 2232 2233 struct device dev; 2234 const struct attribute_group *sysfs_groups[4]; 2235 const struct attribute_group *sysfs_rx_queue_group; 2236 2237 const struct rtnl_link_ops *rtnl_link_ops; 2238 2239 /* for setting kernel sock attribute on TCP connection setup */ 2240 #define GSO_MAX_SIZE 65536 2241 unsigned int gso_max_size; 2242 #define GSO_MAX_SEGS 65535 2243 u16 gso_max_segs; 2244 2245 #ifdef CONFIG_DCB 2246 const struct dcbnl_rtnl_ops *dcbnl_ops; 2247 #endif 2248 s16 num_tc; 2249 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE]; 2250 u8 prio_tc_map[TC_BITMASK + 1]; 2251 2252 #if IS_ENABLED(CONFIG_FCOE) 2253 unsigned int fcoe_ddp_xid; 2254 #endif 2255 #if IS_ENABLED(CONFIG_CGROUP_NET_PRIO) 2256 struct netprio_map __rcu *priomap; 2257 #endif 2258 struct phy_device *phydev; 2259 struct sfp_bus *sfp_bus; 2260 struct lock_class_key *qdisc_tx_busylock; 2261 bool proto_down; 2262 unsigned wol_enabled:1; 2263 unsigned threaded:1; 2264 2265 struct list_head net_notifier_list; 2266 2267 #if IS_ENABLED(CONFIG_MACSEC) 2268 /* MACsec management functions */ 2269 const struct macsec_ops *macsec_ops; 2270 #endif 2271 const struct udp_tunnel_nic_info *udp_tunnel_nic_info; 2272 struct udp_tunnel_nic *udp_tunnel_nic; 2273 2274 /* protected by rtnl_lock */ 2275 struct bpf_xdp_entity xdp_state[__MAX_XDP_MODE]; 2276 2277 u8 dev_addr_shadow[MAX_ADDR_LEN]; 2278 netdevice_tracker linkwatch_dev_tracker; 2279 netdevice_tracker watchdog_dev_tracker; 2280 }; 2281 #define to_net_dev(d) container_of(d, struct net_device, dev) 2282 2283 static inline bool netif_elide_gro(const struct net_device *dev) 2284 { 2285 if (!(dev->features & NETIF_F_GRO) || dev->xdp_prog) 2286 return true; 2287 return false; 2288 } 2289 2290 #define NETDEV_ALIGN 32 2291 2292 static inline 2293 int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio) 2294 { 2295 return dev->prio_tc_map[prio & TC_BITMASK]; 2296 } 2297 2298 static inline 2299 int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc) 2300 { 2301 if (tc >= dev->num_tc) 2302 return -EINVAL; 2303 2304 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK; 2305 return 0; 2306 } 2307 2308 int netdev_txq_to_tc(struct net_device *dev, unsigned int txq); 2309 void netdev_reset_tc(struct net_device *dev); 2310 int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset); 2311 int netdev_set_num_tc(struct net_device *dev, u8 num_tc); 2312 2313 static inline 2314 int netdev_get_num_tc(struct net_device *dev) 2315 { 2316 return dev->num_tc; 2317 } 2318 2319 static inline void net_prefetch(void *p) 2320 { 2321 prefetch(p); 2322 #if L1_CACHE_BYTES < 128 2323 prefetch((u8 *)p + L1_CACHE_BYTES); 2324 #endif 2325 } 2326 2327 static inline void net_prefetchw(void *p) 2328 { 2329 prefetchw(p); 2330 #if L1_CACHE_BYTES < 128 2331 prefetchw((u8 *)p + L1_CACHE_BYTES); 2332 #endif 2333 } 2334 2335 void netdev_unbind_sb_channel(struct net_device *dev, 2336 struct net_device *sb_dev); 2337 int netdev_bind_sb_channel_queue(struct net_device *dev, 2338 struct net_device *sb_dev, 2339 u8 tc, u16 count, u16 offset); 2340 int netdev_set_sb_channel(struct net_device *dev, u16 channel); 2341 static inline int netdev_get_sb_channel(struct net_device *dev) 2342 { 2343 return max_t(int, -dev->num_tc, 0); 2344 } 2345 2346 static inline 2347 struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev, 2348 unsigned int index) 2349 { 2350 return &dev->_tx[index]; 2351 } 2352 2353 static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev, 2354 const struct sk_buff *skb) 2355 { 2356 return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb)); 2357 } 2358 2359 static inline void netdev_for_each_tx_queue(struct net_device *dev, 2360 void (*f)(struct net_device *, 2361 struct netdev_queue *, 2362 void *), 2363 void *arg) 2364 { 2365 unsigned int i; 2366 2367 for (i = 0; i < dev->num_tx_queues; i++) 2368 f(dev, &dev->_tx[i], arg); 2369 } 2370 2371 #define netdev_lockdep_set_classes(dev) \ 2372 { \ 2373 static struct lock_class_key qdisc_tx_busylock_key; \ 2374 static struct lock_class_key qdisc_xmit_lock_key; \ 2375 static struct lock_class_key dev_addr_list_lock_key; \ 2376 unsigned int i; \ 2377 \ 2378 (dev)->qdisc_tx_busylock = &qdisc_tx_busylock_key; \ 2379 lockdep_set_class(&(dev)->addr_list_lock, \ 2380 &dev_addr_list_lock_key); \ 2381 for (i = 0; i < (dev)->num_tx_queues; i++) \ 2382 lockdep_set_class(&(dev)->_tx[i]._xmit_lock, \ 2383 &qdisc_xmit_lock_key); \ 2384 } 2385 2386 u16 netdev_pick_tx(struct net_device *dev, struct sk_buff *skb, 2387 struct net_device *sb_dev); 2388 struct netdev_queue *netdev_core_pick_tx(struct net_device *dev, 2389 struct sk_buff *skb, 2390 struct net_device *sb_dev); 2391 2392 /* returns the headroom that the master device needs to take in account 2393 * when forwarding to this dev 2394 */ 2395 static inline unsigned netdev_get_fwd_headroom(struct net_device *dev) 2396 { 2397 return dev->priv_flags & IFF_PHONY_HEADROOM ? 0 : dev->needed_headroom; 2398 } 2399 2400 static inline void netdev_set_rx_headroom(struct net_device *dev, int new_hr) 2401 { 2402 if (dev->netdev_ops->ndo_set_rx_headroom) 2403 dev->netdev_ops->ndo_set_rx_headroom(dev, new_hr); 2404 } 2405 2406 /* set the device rx headroom to the dev's default */ 2407 static inline void netdev_reset_rx_headroom(struct net_device *dev) 2408 { 2409 netdev_set_rx_headroom(dev, -1); 2410 } 2411 2412 static inline void *netdev_get_ml_priv(struct net_device *dev, 2413 enum netdev_ml_priv_type type) 2414 { 2415 if (dev->ml_priv_type != type) 2416 return NULL; 2417 2418 return dev->ml_priv; 2419 } 2420 2421 static inline void netdev_set_ml_priv(struct net_device *dev, 2422 void *ml_priv, 2423 enum netdev_ml_priv_type type) 2424 { 2425 WARN(dev->ml_priv_type && dev->ml_priv_type != type, 2426 "Overwriting already set ml_priv_type (%u) with different ml_priv_type (%u)!\n", 2427 dev->ml_priv_type, type); 2428 WARN(!dev->ml_priv_type && dev->ml_priv, 2429 "Overwriting already set ml_priv and ml_priv_type is ML_PRIV_NONE!\n"); 2430 2431 dev->ml_priv = ml_priv; 2432 dev->ml_priv_type = type; 2433 } 2434 2435 /* 2436 * Net namespace inlines 2437 */ 2438 static inline 2439 struct net *dev_net(const struct net_device *dev) 2440 { 2441 return read_pnet(&dev->nd_net); 2442 } 2443 2444 static inline 2445 void dev_net_set(struct net_device *dev, struct net *net) 2446 { 2447 write_pnet(&dev->nd_net, net); 2448 } 2449 2450 /** 2451 * netdev_priv - access network device private data 2452 * @dev: network device 2453 * 2454 * Get network device private data 2455 */ 2456 static inline void *netdev_priv(const struct net_device *dev) 2457 { 2458 return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN); 2459 } 2460 2461 /* Set the sysfs physical device reference for the network logical device 2462 * if set prior to registration will cause a symlink during initialization. 2463 */ 2464 #define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev)) 2465 2466 /* Set the sysfs device type for the network logical device to allow 2467 * fine-grained identification of different network device types. For 2468 * example Ethernet, Wireless LAN, Bluetooth, WiMAX etc. 2469 */ 2470 #define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype)) 2471 2472 /* Default NAPI poll() weight 2473 * Device drivers are strongly advised to not use bigger value 2474 */ 2475 #define NAPI_POLL_WEIGHT 64 2476 2477 /** 2478 * netif_napi_add - initialize a NAPI context 2479 * @dev: network device 2480 * @napi: NAPI context 2481 * @poll: polling function 2482 * @weight: default weight 2483 * 2484 * netif_napi_add() must be used to initialize a NAPI context prior to calling 2485 * *any* of the other NAPI-related functions. 2486 */ 2487 void netif_napi_add(struct net_device *dev, struct napi_struct *napi, 2488 int (*poll)(struct napi_struct *, int), int weight); 2489 2490 /** 2491 * netif_tx_napi_add - initialize a NAPI context 2492 * @dev: network device 2493 * @napi: NAPI context 2494 * @poll: polling function 2495 * @weight: default weight 2496 * 2497 * This variant of netif_napi_add() should be used from drivers using NAPI 2498 * to exclusively poll a TX queue. 2499 * This will avoid we add it into napi_hash[], thus polluting this hash table. 2500 */ 2501 static inline void netif_tx_napi_add(struct net_device *dev, 2502 struct napi_struct *napi, 2503 int (*poll)(struct napi_struct *, int), 2504 int weight) 2505 { 2506 set_bit(NAPI_STATE_NO_BUSY_POLL, &napi->state); 2507 netif_napi_add(dev, napi, poll, weight); 2508 } 2509 2510 /** 2511 * __netif_napi_del - remove a NAPI context 2512 * @napi: NAPI context 2513 * 2514 * Warning: caller must observe RCU grace period before freeing memory 2515 * containing @napi. Drivers might want to call this helper to combine 2516 * all the needed RCU grace periods into a single one. 2517 */ 2518 void __netif_napi_del(struct napi_struct *napi); 2519 2520 /** 2521 * netif_napi_del - remove a NAPI context 2522 * @napi: NAPI context 2523 * 2524 * netif_napi_del() removes a NAPI context from the network device NAPI list 2525 */ 2526 static inline void netif_napi_del(struct napi_struct *napi) 2527 { 2528 __netif_napi_del(napi); 2529 synchronize_net(); 2530 } 2531 2532 struct packet_type { 2533 __be16 type; /* This is really htons(ether_type). */ 2534 bool ignore_outgoing; 2535 struct net_device *dev; /* NULL is wildcarded here */ 2536 netdevice_tracker dev_tracker; 2537 int (*func) (struct sk_buff *, 2538 struct net_device *, 2539 struct packet_type *, 2540 struct net_device *); 2541 void (*list_func) (struct list_head *, 2542 struct packet_type *, 2543 struct net_device *); 2544 bool (*id_match)(struct packet_type *ptype, 2545 struct sock *sk); 2546 void *af_packet_priv; 2547 struct list_head list; 2548 }; 2549 2550 struct offload_callbacks { 2551 struct sk_buff *(*gso_segment)(struct sk_buff *skb, 2552 netdev_features_t features); 2553 struct sk_buff *(*gro_receive)(struct list_head *head, 2554 struct sk_buff *skb); 2555 int (*gro_complete)(struct sk_buff *skb, int nhoff); 2556 }; 2557 2558 struct packet_offload { 2559 __be16 type; /* This is really htons(ether_type). */ 2560 u16 priority; 2561 struct offload_callbacks callbacks; 2562 struct list_head list; 2563 }; 2564 2565 /* often modified stats are per-CPU, other are shared (netdev->stats) */ 2566 struct pcpu_sw_netstats { 2567 u64 rx_packets; 2568 u64 rx_bytes; 2569 u64 tx_packets; 2570 u64 tx_bytes; 2571 struct u64_stats_sync syncp; 2572 } __aligned(4 * sizeof(u64)); 2573 2574 struct pcpu_lstats { 2575 u64_stats_t packets; 2576 u64_stats_t bytes; 2577 struct u64_stats_sync syncp; 2578 } __aligned(2 * sizeof(u64)); 2579 2580 void dev_lstats_read(struct net_device *dev, u64 *packets, u64 *bytes); 2581 2582 static inline void dev_sw_netstats_rx_add(struct net_device *dev, unsigned int len) 2583 { 2584 struct pcpu_sw_netstats *tstats = this_cpu_ptr(dev->tstats); 2585 2586 u64_stats_update_begin(&tstats->syncp); 2587 tstats->rx_bytes += len; 2588 tstats->rx_packets++; 2589 u64_stats_update_end(&tstats->syncp); 2590 } 2591 2592 static inline void dev_sw_netstats_tx_add(struct net_device *dev, 2593 unsigned int packets, 2594 unsigned int len) 2595 { 2596 struct pcpu_sw_netstats *tstats = this_cpu_ptr(dev->tstats); 2597 2598 u64_stats_update_begin(&tstats->syncp); 2599 tstats->tx_bytes += len; 2600 tstats->tx_packets += packets; 2601 u64_stats_update_end(&tstats->syncp); 2602 } 2603 2604 static inline void dev_lstats_add(struct net_device *dev, unsigned int len) 2605 { 2606 struct pcpu_lstats *lstats = this_cpu_ptr(dev->lstats); 2607 2608 u64_stats_update_begin(&lstats->syncp); 2609 u64_stats_add(&lstats->bytes, len); 2610 u64_stats_inc(&lstats->packets); 2611 u64_stats_update_end(&lstats->syncp); 2612 } 2613 2614 #define __netdev_alloc_pcpu_stats(type, gfp) \ 2615 ({ \ 2616 typeof(type) __percpu *pcpu_stats = alloc_percpu_gfp(type, gfp);\ 2617 if (pcpu_stats) { \ 2618 int __cpu; \ 2619 for_each_possible_cpu(__cpu) { \ 2620 typeof(type) *stat; \ 2621 stat = per_cpu_ptr(pcpu_stats, __cpu); \ 2622 u64_stats_init(&stat->syncp); \ 2623 } \ 2624 } \ 2625 pcpu_stats; \ 2626 }) 2627 2628 #define netdev_alloc_pcpu_stats(type) \ 2629 __netdev_alloc_pcpu_stats(type, GFP_KERNEL) 2630 2631 #define devm_netdev_alloc_pcpu_stats(dev, type) \ 2632 ({ \ 2633 typeof(type) __percpu *pcpu_stats = devm_alloc_percpu(dev, type);\ 2634 if (pcpu_stats) { \ 2635 int __cpu; \ 2636 for_each_possible_cpu(__cpu) { \ 2637 typeof(type) *stat; \ 2638 stat = per_cpu_ptr(pcpu_stats, __cpu); \ 2639 u64_stats_init(&stat->syncp); \ 2640 } \ 2641 } \ 2642 pcpu_stats; \ 2643 }) 2644 2645 enum netdev_lag_tx_type { 2646 NETDEV_LAG_TX_TYPE_UNKNOWN, 2647 NETDEV_LAG_TX_TYPE_RANDOM, 2648 NETDEV_LAG_TX_TYPE_BROADCAST, 2649 NETDEV_LAG_TX_TYPE_ROUNDROBIN, 2650 NETDEV_LAG_TX_TYPE_ACTIVEBACKUP, 2651 NETDEV_LAG_TX_TYPE_HASH, 2652 }; 2653 2654 enum netdev_lag_hash { 2655 NETDEV_LAG_HASH_NONE, 2656 NETDEV_LAG_HASH_L2, 2657 NETDEV_LAG_HASH_L34, 2658 NETDEV_LAG_HASH_L23, 2659 NETDEV_LAG_HASH_E23, 2660 NETDEV_LAG_HASH_E34, 2661 NETDEV_LAG_HASH_VLAN_SRCMAC, 2662 NETDEV_LAG_HASH_UNKNOWN, 2663 }; 2664 2665 struct netdev_lag_upper_info { 2666 enum netdev_lag_tx_type tx_type; 2667 enum netdev_lag_hash hash_type; 2668 }; 2669 2670 struct netdev_lag_lower_state_info { 2671 u8 link_up : 1, 2672 tx_enabled : 1; 2673 }; 2674 2675 #include <linux/notifier.h> 2676 2677 /* netdevice notifier chain. Please remember to update netdev_cmd_to_name() 2678 * and the rtnetlink notification exclusion list in rtnetlink_event() when 2679 * adding new types. 2680 */ 2681 enum netdev_cmd { 2682 NETDEV_UP = 1, /* For now you can't veto a device up/down */ 2683 NETDEV_DOWN, 2684 NETDEV_REBOOT, /* Tell a protocol stack a network interface 2685 detected a hardware crash and restarted 2686 - we can use this eg to kick tcp sessions 2687 once done */ 2688 NETDEV_CHANGE, /* Notify device state change */ 2689 NETDEV_REGISTER, 2690 NETDEV_UNREGISTER, 2691 NETDEV_CHANGEMTU, /* notify after mtu change happened */ 2692 NETDEV_CHANGEADDR, /* notify after the address change */ 2693 NETDEV_PRE_CHANGEADDR, /* notify before the address change */ 2694 NETDEV_GOING_DOWN, 2695 NETDEV_CHANGENAME, 2696 NETDEV_FEAT_CHANGE, 2697 NETDEV_BONDING_FAILOVER, 2698 NETDEV_PRE_UP, 2699 NETDEV_PRE_TYPE_CHANGE, 2700 NETDEV_POST_TYPE_CHANGE, 2701 NETDEV_POST_INIT, 2702 NETDEV_RELEASE, 2703 NETDEV_NOTIFY_PEERS, 2704 NETDEV_JOIN, 2705 NETDEV_CHANGEUPPER, 2706 NETDEV_RESEND_IGMP, 2707 NETDEV_PRECHANGEMTU, /* notify before mtu change happened */ 2708 NETDEV_CHANGEINFODATA, 2709 NETDEV_BONDING_INFO, 2710 NETDEV_PRECHANGEUPPER, 2711 NETDEV_CHANGELOWERSTATE, 2712 NETDEV_UDP_TUNNEL_PUSH_INFO, 2713 NETDEV_UDP_TUNNEL_DROP_INFO, 2714 NETDEV_CHANGE_TX_QUEUE_LEN, 2715 NETDEV_CVLAN_FILTER_PUSH_INFO, 2716 NETDEV_CVLAN_FILTER_DROP_INFO, 2717 NETDEV_SVLAN_FILTER_PUSH_INFO, 2718 NETDEV_SVLAN_FILTER_DROP_INFO, 2719 }; 2720 const char *netdev_cmd_to_name(enum netdev_cmd cmd); 2721 2722 int register_netdevice_notifier(struct notifier_block *nb); 2723 int unregister_netdevice_notifier(struct notifier_block *nb); 2724 int register_netdevice_notifier_net(struct net *net, struct notifier_block *nb); 2725 int unregister_netdevice_notifier_net(struct net *net, 2726 struct notifier_block *nb); 2727 int register_netdevice_notifier_dev_net(struct net_device *dev, 2728 struct notifier_block *nb, 2729 struct netdev_net_notifier *nn); 2730 int unregister_netdevice_notifier_dev_net(struct net_device *dev, 2731 struct notifier_block *nb, 2732 struct netdev_net_notifier *nn); 2733 2734 struct netdev_notifier_info { 2735 struct net_device *dev; 2736 struct netlink_ext_ack *extack; 2737 }; 2738 2739 struct netdev_notifier_info_ext { 2740 struct netdev_notifier_info info; /* must be first */ 2741 union { 2742 u32 mtu; 2743 } ext; 2744 }; 2745 2746 struct netdev_notifier_change_info { 2747 struct netdev_notifier_info info; /* must be first */ 2748 unsigned int flags_changed; 2749 }; 2750 2751 struct netdev_notifier_changeupper_info { 2752 struct netdev_notifier_info info; /* must be first */ 2753 struct net_device *upper_dev; /* new upper dev */ 2754 bool master; /* is upper dev master */ 2755 bool linking; /* is the notification for link or unlink */ 2756 void *upper_info; /* upper dev info */ 2757 }; 2758 2759 struct netdev_notifier_changelowerstate_info { 2760 struct netdev_notifier_info info; /* must be first */ 2761 void *lower_state_info; /* is lower dev state */ 2762 }; 2763 2764 struct netdev_notifier_pre_changeaddr_info { 2765 struct netdev_notifier_info info; /* must be first */ 2766 const unsigned char *dev_addr; 2767 }; 2768 2769 static inline void netdev_notifier_info_init(struct netdev_notifier_info *info, 2770 struct net_device *dev) 2771 { 2772 info->dev = dev; 2773 info->extack = NULL; 2774 } 2775 2776 static inline struct net_device * 2777 netdev_notifier_info_to_dev(const struct netdev_notifier_info *info) 2778 { 2779 return info->dev; 2780 } 2781 2782 static inline struct netlink_ext_ack * 2783 netdev_notifier_info_to_extack(const struct netdev_notifier_info *info) 2784 { 2785 return info->extack; 2786 } 2787 2788 int call_netdevice_notifiers(unsigned long val, struct net_device *dev); 2789 2790 2791 extern rwlock_t dev_base_lock; /* Device list lock */ 2792 2793 #define for_each_netdev(net, d) \ 2794 list_for_each_entry(d, &(net)->dev_base_head, dev_list) 2795 #define for_each_netdev_reverse(net, d) \ 2796 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list) 2797 #define for_each_netdev_rcu(net, d) \ 2798 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list) 2799 #define for_each_netdev_safe(net, d, n) \ 2800 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list) 2801 #define for_each_netdev_continue(net, d) \ 2802 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list) 2803 #define for_each_netdev_continue_reverse(net, d) \ 2804 list_for_each_entry_continue_reverse(d, &(net)->dev_base_head, \ 2805 dev_list) 2806 #define for_each_netdev_continue_rcu(net, d) \ 2807 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list) 2808 #define for_each_netdev_in_bond_rcu(bond, slave) \ 2809 for_each_netdev_rcu(&init_net, slave) \ 2810 if (netdev_master_upper_dev_get_rcu(slave) == (bond)) 2811 #define net_device_entry(lh) list_entry(lh, struct net_device, dev_list) 2812 2813 static inline struct net_device *next_net_device(struct net_device *dev) 2814 { 2815 struct list_head *lh; 2816 struct net *net; 2817 2818 net = dev_net(dev); 2819 lh = dev->dev_list.next; 2820 return lh == &net->dev_base_head ? NULL : net_device_entry(lh); 2821 } 2822 2823 static inline struct net_device *next_net_device_rcu(struct net_device *dev) 2824 { 2825 struct list_head *lh; 2826 struct net *net; 2827 2828 net = dev_net(dev); 2829 lh = rcu_dereference(list_next_rcu(&dev->dev_list)); 2830 return lh == &net->dev_base_head ? NULL : net_device_entry(lh); 2831 } 2832 2833 static inline struct net_device *first_net_device(struct net *net) 2834 { 2835 return list_empty(&net->dev_base_head) ? NULL : 2836 net_device_entry(net->dev_base_head.next); 2837 } 2838 2839 static inline struct net_device *first_net_device_rcu(struct net *net) 2840 { 2841 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head)); 2842 2843 return lh == &net->dev_base_head ? NULL : net_device_entry(lh); 2844 } 2845 2846 int netdev_boot_setup_check(struct net_device *dev); 2847 struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type, 2848 const char *hwaddr); 2849 struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type); 2850 void dev_add_pack(struct packet_type *pt); 2851 void dev_remove_pack(struct packet_type *pt); 2852 void __dev_remove_pack(struct packet_type *pt); 2853 void dev_add_offload(struct packet_offload *po); 2854 void dev_remove_offload(struct packet_offload *po); 2855 2856 int dev_get_iflink(const struct net_device *dev); 2857 int dev_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb); 2858 int dev_fill_forward_path(const struct net_device *dev, const u8 *daddr, 2859 struct net_device_path_stack *stack); 2860 struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags, 2861 unsigned short mask); 2862 struct net_device *dev_get_by_name(struct net *net, const char *name); 2863 struct net_device *dev_get_by_name_rcu(struct net *net, const char *name); 2864 struct net_device *__dev_get_by_name(struct net *net, const char *name); 2865 bool netdev_name_in_use(struct net *net, const char *name); 2866 int dev_alloc_name(struct net_device *dev, const char *name); 2867 int dev_open(struct net_device *dev, struct netlink_ext_ack *extack); 2868 void dev_close(struct net_device *dev); 2869 void dev_close_many(struct list_head *head, bool unlink); 2870 void dev_disable_lro(struct net_device *dev); 2871 int dev_loopback_xmit(struct net *net, struct sock *sk, struct sk_buff *newskb); 2872 u16 dev_pick_tx_zero(struct net_device *dev, struct sk_buff *skb, 2873 struct net_device *sb_dev); 2874 u16 dev_pick_tx_cpu_id(struct net_device *dev, struct sk_buff *skb, 2875 struct net_device *sb_dev); 2876 2877 int dev_queue_xmit(struct sk_buff *skb); 2878 int dev_queue_xmit_accel(struct sk_buff *skb, struct net_device *sb_dev); 2879 int __dev_direct_xmit(struct sk_buff *skb, u16 queue_id); 2880 2881 static inline int dev_direct_xmit(struct sk_buff *skb, u16 queue_id) 2882 { 2883 int ret; 2884 2885 ret = __dev_direct_xmit(skb, queue_id); 2886 if (!dev_xmit_complete(ret)) 2887 kfree_skb(skb); 2888 return ret; 2889 } 2890 2891 int register_netdevice(struct net_device *dev); 2892 void unregister_netdevice_queue(struct net_device *dev, struct list_head *head); 2893 void unregister_netdevice_many(struct list_head *head); 2894 static inline void unregister_netdevice(struct net_device *dev) 2895 { 2896 unregister_netdevice_queue(dev, NULL); 2897 } 2898 2899 int netdev_refcnt_read(const struct net_device *dev); 2900 void free_netdev(struct net_device *dev); 2901 void netdev_freemem(struct net_device *dev); 2902 int init_dummy_netdev(struct net_device *dev); 2903 2904 struct net_device *netdev_get_xmit_slave(struct net_device *dev, 2905 struct sk_buff *skb, 2906 bool all_slaves); 2907 struct net_device *netdev_sk_get_lowest_dev(struct net_device *dev, 2908 struct sock *sk); 2909 struct net_device *dev_get_by_index(struct net *net, int ifindex); 2910 struct net_device *__dev_get_by_index(struct net *net, int ifindex); 2911 struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex); 2912 struct net_device *dev_get_by_napi_id(unsigned int napi_id); 2913 int netdev_get_name(struct net *net, char *name, int ifindex); 2914 int dev_restart(struct net_device *dev); 2915 2916 2917 static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev, 2918 unsigned short type, 2919 const void *daddr, const void *saddr, 2920 unsigned int len) 2921 { 2922 if (!dev->header_ops || !dev->header_ops->create) 2923 return 0; 2924 2925 return dev->header_ops->create(skb, dev, type, daddr, saddr, len); 2926 } 2927 2928 static inline int dev_parse_header(const struct sk_buff *skb, 2929 unsigned char *haddr) 2930 { 2931 const struct net_device *dev = skb->dev; 2932 2933 if (!dev->header_ops || !dev->header_ops->parse) 2934 return 0; 2935 return dev->header_ops->parse(skb, haddr); 2936 } 2937 2938 static inline __be16 dev_parse_header_protocol(const struct sk_buff *skb) 2939 { 2940 const struct net_device *dev = skb->dev; 2941 2942 if (!dev->header_ops || !dev->header_ops->parse_protocol) 2943 return 0; 2944 return dev->header_ops->parse_protocol(skb); 2945 } 2946 2947 /* ll_header must have at least hard_header_len allocated */ 2948 static inline bool dev_validate_header(const struct net_device *dev, 2949 char *ll_header, int len) 2950 { 2951 if (likely(len >= dev->hard_header_len)) 2952 return true; 2953 if (len < dev->min_header_len) 2954 return false; 2955 2956 if (capable(CAP_SYS_RAWIO)) { 2957 memset(ll_header + len, 0, dev->hard_header_len - len); 2958 return true; 2959 } 2960 2961 if (dev->header_ops && dev->header_ops->validate) 2962 return dev->header_ops->validate(ll_header, len); 2963 2964 return false; 2965 } 2966 2967 static inline bool dev_has_header(const struct net_device *dev) 2968 { 2969 return dev->header_ops && dev->header_ops->create; 2970 } 2971 2972 #ifdef CONFIG_NET_FLOW_LIMIT 2973 #define FLOW_LIMIT_HISTORY (1 << 7) /* must be ^2 and !overflow buckets */ 2974 struct sd_flow_limit { 2975 u64 count; 2976 unsigned int num_buckets; 2977 unsigned int history_head; 2978 u16 history[FLOW_LIMIT_HISTORY]; 2979 u8 buckets[]; 2980 }; 2981 2982 extern int netdev_flow_limit_table_len; 2983 #endif /* CONFIG_NET_FLOW_LIMIT */ 2984 2985 /* 2986 * Incoming packets are placed on per-CPU queues 2987 */ 2988 struct softnet_data { 2989 struct list_head poll_list; 2990 struct sk_buff_head process_queue; 2991 2992 /* stats */ 2993 unsigned int processed; 2994 unsigned int time_squeeze; 2995 unsigned int received_rps; 2996 #ifdef CONFIG_RPS 2997 struct softnet_data *rps_ipi_list; 2998 #endif 2999 #ifdef CONFIG_NET_FLOW_LIMIT 3000 struct sd_flow_limit __rcu *flow_limit; 3001 #endif 3002 struct Qdisc *output_queue; 3003 struct Qdisc **output_queue_tailp; 3004 struct sk_buff *completion_queue; 3005 #ifdef CONFIG_XFRM_OFFLOAD 3006 struct sk_buff_head xfrm_backlog; 3007 #endif 3008 /* written and read only by owning cpu: */ 3009 struct { 3010 u16 recursion; 3011 u8 more; 3012 } xmit; 3013 #ifdef CONFIG_RPS 3014 /* input_queue_head should be written by cpu owning this struct, 3015 * and only read by other cpus. Worth using a cache line. 3016 */ 3017 unsigned int input_queue_head ____cacheline_aligned_in_smp; 3018 3019 /* Elements below can be accessed between CPUs for RPS/RFS */ 3020 call_single_data_t csd ____cacheline_aligned_in_smp; 3021 struct softnet_data *rps_ipi_next; 3022 unsigned int cpu; 3023 unsigned int input_queue_tail; 3024 #endif 3025 unsigned int dropped; 3026 struct sk_buff_head input_pkt_queue; 3027 struct napi_struct backlog; 3028 3029 }; 3030 3031 static inline void input_queue_head_incr(struct softnet_data *sd) 3032 { 3033 #ifdef CONFIG_RPS 3034 sd->input_queue_head++; 3035 #endif 3036 } 3037 3038 static inline void input_queue_tail_incr_save(struct softnet_data *sd, 3039 unsigned int *qtail) 3040 { 3041 #ifdef CONFIG_RPS 3042 *qtail = ++sd->input_queue_tail; 3043 #endif 3044 } 3045 3046 DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data); 3047 3048 static inline int dev_recursion_level(void) 3049 { 3050 return this_cpu_read(softnet_data.xmit.recursion); 3051 } 3052 3053 #define XMIT_RECURSION_LIMIT 8 3054 static inline bool dev_xmit_recursion(void) 3055 { 3056 return unlikely(__this_cpu_read(softnet_data.xmit.recursion) > 3057 XMIT_RECURSION_LIMIT); 3058 } 3059 3060 static inline void dev_xmit_recursion_inc(void) 3061 { 3062 __this_cpu_inc(softnet_data.xmit.recursion); 3063 } 3064 3065 static inline void dev_xmit_recursion_dec(void) 3066 { 3067 __this_cpu_dec(softnet_data.xmit.recursion); 3068 } 3069 3070 void __netif_schedule(struct Qdisc *q); 3071 void netif_schedule_queue(struct netdev_queue *txq); 3072 3073 static inline void netif_tx_schedule_all(struct net_device *dev) 3074 { 3075 unsigned int i; 3076 3077 for (i = 0; i < dev->num_tx_queues; i++) 3078 netif_schedule_queue(netdev_get_tx_queue(dev, i)); 3079 } 3080 3081 static __always_inline void netif_tx_start_queue(struct netdev_queue *dev_queue) 3082 { 3083 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state); 3084 } 3085 3086 /** 3087 * netif_start_queue - allow transmit 3088 * @dev: network device 3089 * 3090 * Allow upper layers to call the device hard_start_xmit routine. 3091 */ 3092 static inline void netif_start_queue(struct net_device *dev) 3093 { 3094 netif_tx_start_queue(netdev_get_tx_queue(dev, 0)); 3095 } 3096 3097 static inline void netif_tx_start_all_queues(struct net_device *dev) 3098 { 3099 unsigned int i; 3100 3101 for (i = 0; i < dev->num_tx_queues; i++) { 3102 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 3103 netif_tx_start_queue(txq); 3104 } 3105 } 3106 3107 void netif_tx_wake_queue(struct netdev_queue *dev_queue); 3108 3109 /** 3110 * netif_wake_queue - restart transmit 3111 * @dev: network device 3112 * 3113 * Allow upper layers to call the device hard_start_xmit routine. 3114 * Used for flow control when transmit resources are available. 3115 */ 3116 static inline void netif_wake_queue(struct net_device *dev) 3117 { 3118 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0)); 3119 } 3120 3121 static inline void netif_tx_wake_all_queues(struct net_device *dev) 3122 { 3123 unsigned int i; 3124 3125 for (i = 0; i < dev->num_tx_queues; i++) { 3126 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 3127 netif_tx_wake_queue(txq); 3128 } 3129 } 3130 3131 static __always_inline void netif_tx_stop_queue(struct netdev_queue *dev_queue) 3132 { 3133 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state); 3134 } 3135 3136 /** 3137 * netif_stop_queue - stop transmitted packets 3138 * @dev: network device 3139 * 3140 * Stop upper layers calling the device hard_start_xmit routine. 3141 * Used for flow control when transmit resources are unavailable. 3142 */ 3143 static inline void netif_stop_queue(struct net_device *dev) 3144 { 3145 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0)); 3146 } 3147 3148 void netif_tx_stop_all_queues(struct net_device *dev); 3149 3150 static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue) 3151 { 3152 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state); 3153 } 3154 3155 /** 3156 * netif_queue_stopped - test if transmit queue is flowblocked 3157 * @dev: network device 3158 * 3159 * Test if transmit queue on device is currently unable to send. 3160 */ 3161 static inline bool netif_queue_stopped(const struct net_device *dev) 3162 { 3163 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0)); 3164 } 3165 3166 static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue) 3167 { 3168 return dev_queue->state & QUEUE_STATE_ANY_XOFF; 3169 } 3170 3171 static inline bool 3172 netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue) 3173 { 3174 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN; 3175 } 3176 3177 static inline bool 3178 netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue) 3179 { 3180 return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN; 3181 } 3182 3183 /** 3184 * netdev_queue_set_dql_min_limit - set dql minimum limit 3185 * @dev_queue: pointer to transmit queue 3186 * @min_limit: dql minimum limit 3187 * 3188 * Forces xmit_more() to return true until the minimum threshold 3189 * defined by @min_limit is reached (or until the tx queue is 3190 * empty). Warning: to be use with care, misuse will impact the 3191 * latency. 3192 */ 3193 static inline void netdev_queue_set_dql_min_limit(struct netdev_queue *dev_queue, 3194 unsigned int min_limit) 3195 { 3196 #ifdef CONFIG_BQL 3197 dev_queue->dql.min_limit = min_limit; 3198 #endif 3199 } 3200 3201 /** 3202 * netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write 3203 * @dev_queue: pointer to transmit queue 3204 * 3205 * BQL enabled drivers might use this helper in their ndo_start_xmit(), 3206 * to give appropriate hint to the CPU. 3207 */ 3208 static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue) 3209 { 3210 #ifdef CONFIG_BQL 3211 prefetchw(&dev_queue->dql.num_queued); 3212 #endif 3213 } 3214 3215 /** 3216 * netdev_txq_bql_complete_prefetchw - prefetch bql data for write 3217 * @dev_queue: pointer to transmit queue 3218 * 3219 * BQL enabled drivers might use this helper in their TX completion path, 3220 * to give appropriate hint to the CPU. 3221 */ 3222 static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue) 3223 { 3224 #ifdef CONFIG_BQL 3225 prefetchw(&dev_queue->dql.limit); 3226 #endif 3227 } 3228 3229 static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue, 3230 unsigned int bytes) 3231 { 3232 #ifdef CONFIG_BQL 3233 dql_queued(&dev_queue->dql, bytes); 3234 3235 if (likely(dql_avail(&dev_queue->dql) >= 0)) 3236 return; 3237 3238 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state); 3239 3240 /* 3241 * The XOFF flag must be set before checking the dql_avail below, 3242 * because in netdev_tx_completed_queue we update the dql_completed 3243 * before checking the XOFF flag. 3244 */ 3245 smp_mb(); 3246 3247 /* check again in case another CPU has just made room avail */ 3248 if (unlikely(dql_avail(&dev_queue->dql) >= 0)) 3249 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state); 3250 #endif 3251 } 3252 3253 /* Variant of netdev_tx_sent_queue() for drivers that are aware 3254 * that they should not test BQL status themselves. 3255 * We do want to change __QUEUE_STATE_STACK_XOFF only for the last 3256 * skb of a batch. 3257 * Returns true if the doorbell must be used to kick the NIC. 3258 */ 3259 static inline bool __netdev_tx_sent_queue(struct netdev_queue *dev_queue, 3260 unsigned int bytes, 3261 bool xmit_more) 3262 { 3263 if (xmit_more) { 3264 #ifdef CONFIG_BQL 3265 dql_queued(&dev_queue->dql, bytes); 3266 #endif 3267 return netif_tx_queue_stopped(dev_queue); 3268 } 3269 netdev_tx_sent_queue(dev_queue, bytes); 3270 return true; 3271 } 3272 3273 /** 3274 * netdev_sent_queue - report the number of bytes queued to hardware 3275 * @dev: network device 3276 * @bytes: number of bytes queued to the hardware device queue 3277 * 3278 * Report the number of bytes queued for sending/completion to the network 3279 * device hardware queue. @bytes should be a good approximation and should 3280 * exactly match netdev_completed_queue() @bytes 3281 */ 3282 static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes) 3283 { 3284 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes); 3285 } 3286 3287 static inline bool __netdev_sent_queue(struct net_device *dev, 3288 unsigned int bytes, 3289 bool xmit_more) 3290 { 3291 return __netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes, 3292 xmit_more); 3293 } 3294 3295 static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue, 3296 unsigned int pkts, unsigned int bytes) 3297 { 3298 #ifdef CONFIG_BQL 3299 if (unlikely(!bytes)) 3300 return; 3301 3302 dql_completed(&dev_queue->dql, bytes); 3303 3304 /* 3305 * Without the memory barrier there is a small possiblity that 3306 * netdev_tx_sent_queue will miss the update and cause the queue to 3307 * be stopped forever 3308 */ 3309 smp_mb(); 3310 3311 if (unlikely(dql_avail(&dev_queue->dql) < 0)) 3312 return; 3313 3314 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state)) 3315 netif_schedule_queue(dev_queue); 3316 #endif 3317 } 3318 3319 /** 3320 * netdev_completed_queue - report bytes and packets completed by device 3321 * @dev: network device 3322 * @pkts: actual number of packets sent over the medium 3323 * @bytes: actual number of bytes sent over the medium 3324 * 3325 * Report the number of bytes and packets transmitted by the network device 3326 * hardware queue over the physical medium, @bytes must exactly match the 3327 * @bytes amount passed to netdev_sent_queue() 3328 */ 3329 static inline void netdev_completed_queue(struct net_device *dev, 3330 unsigned int pkts, unsigned int bytes) 3331 { 3332 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes); 3333 } 3334 3335 static inline void netdev_tx_reset_queue(struct netdev_queue *q) 3336 { 3337 #ifdef CONFIG_BQL 3338 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state); 3339 dql_reset(&q->dql); 3340 #endif 3341 } 3342 3343 /** 3344 * netdev_reset_queue - reset the packets and bytes count of a network device 3345 * @dev_queue: network device 3346 * 3347 * Reset the bytes and packet count of a network device and clear the 3348 * software flow control OFF bit for this network device 3349 */ 3350 static inline void netdev_reset_queue(struct net_device *dev_queue) 3351 { 3352 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0)); 3353 } 3354 3355 /** 3356 * netdev_cap_txqueue - check if selected tx queue exceeds device queues 3357 * @dev: network device 3358 * @queue_index: given tx queue index 3359 * 3360 * Returns 0 if given tx queue index >= number of device tx queues, 3361 * otherwise returns the originally passed tx queue index. 3362 */ 3363 static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index) 3364 { 3365 if (unlikely(queue_index >= dev->real_num_tx_queues)) { 3366 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n", 3367 dev->name, queue_index, 3368 dev->real_num_tx_queues); 3369 return 0; 3370 } 3371 3372 return queue_index; 3373 } 3374 3375 /** 3376 * netif_running - test if up 3377 * @dev: network device 3378 * 3379 * Test if the device has been brought up. 3380 */ 3381 static inline bool netif_running(const struct net_device *dev) 3382 { 3383 return test_bit(__LINK_STATE_START, &dev->state); 3384 } 3385 3386 /* 3387 * Routines to manage the subqueues on a device. We only need start, 3388 * stop, and a check if it's stopped. All other device management is 3389 * done at the overall netdevice level. 3390 * Also test the device if we're multiqueue. 3391 */ 3392 3393 /** 3394 * netif_start_subqueue - allow sending packets on subqueue 3395 * @dev: network device 3396 * @queue_index: sub queue index 3397 * 3398 * Start individual transmit queue of a device with multiple transmit queues. 3399 */ 3400 static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index) 3401 { 3402 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 3403 3404 netif_tx_start_queue(txq); 3405 } 3406 3407 /** 3408 * netif_stop_subqueue - stop sending packets on subqueue 3409 * @dev: network device 3410 * @queue_index: sub queue index 3411 * 3412 * Stop individual transmit queue of a device with multiple transmit queues. 3413 */ 3414 static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index) 3415 { 3416 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 3417 netif_tx_stop_queue(txq); 3418 } 3419 3420 /** 3421 * __netif_subqueue_stopped - test status of subqueue 3422 * @dev: network device 3423 * @queue_index: sub queue index 3424 * 3425 * Check individual transmit queue of a device with multiple transmit queues. 3426 */ 3427 static inline bool __netif_subqueue_stopped(const struct net_device *dev, 3428 u16 queue_index) 3429 { 3430 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 3431 3432 return netif_tx_queue_stopped(txq); 3433 } 3434 3435 /** 3436 * netif_subqueue_stopped - test status of subqueue 3437 * @dev: network device 3438 * @skb: sub queue buffer pointer 3439 * 3440 * Check individual transmit queue of a device with multiple transmit queues. 3441 */ 3442 static inline bool netif_subqueue_stopped(const struct net_device *dev, 3443 struct sk_buff *skb) 3444 { 3445 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb)); 3446 } 3447 3448 /** 3449 * netif_wake_subqueue - allow sending packets on subqueue 3450 * @dev: network device 3451 * @queue_index: sub queue index 3452 * 3453 * Resume individual transmit queue of a device with multiple transmit queues. 3454 */ 3455 static inline void netif_wake_subqueue(struct net_device *dev, u16 queue_index) 3456 { 3457 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 3458 3459 netif_tx_wake_queue(txq); 3460 } 3461 3462 #ifdef CONFIG_XPS 3463 int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask, 3464 u16 index); 3465 int __netif_set_xps_queue(struct net_device *dev, const unsigned long *mask, 3466 u16 index, enum xps_map_type type); 3467 3468 /** 3469 * netif_attr_test_mask - Test a CPU or Rx queue set in a mask 3470 * @j: CPU/Rx queue index 3471 * @mask: bitmask of all cpus/rx queues 3472 * @nr_bits: number of bits in the bitmask 3473 * 3474 * Test if a CPU or Rx queue index is set in a mask of all CPU/Rx queues. 3475 */ 3476 static inline bool netif_attr_test_mask(unsigned long j, 3477 const unsigned long *mask, 3478 unsigned int nr_bits) 3479 { 3480 cpu_max_bits_warn(j, nr_bits); 3481 return test_bit(j, mask); 3482 } 3483 3484 /** 3485 * netif_attr_test_online - Test for online CPU/Rx queue 3486 * @j: CPU/Rx queue index 3487 * @online_mask: bitmask for CPUs/Rx queues that are online 3488 * @nr_bits: number of bits in the bitmask 3489 * 3490 * Returns true if a CPU/Rx queue is online. 3491 */ 3492 static inline bool netif_attr_test_online(unsigned long j, 3493 const unsigned long *online_mask, 3494 unsigned int nr_bits) 3495 { 3496 cpu_max_bits_warn(j, nr_bits); 3497 3498 if (online_mask) 3499 return test_bit(j, online_mask); 3500 3501 return (j < nr_bits); 3502 } 3503 3504 /** 3505 * netif_attrmask_next - get the next CPU/Rx queue in a cpu/Rx queues mask 3506 * @n: CPU/Rx queue index 3507 * @srcp: the cpumask/Rx queue mask pointer 3508 * @nr_bits: number of bits in the bitmask 3509 * 3510 * Returns >= nr_bits if no further CPUs/Rx queues set. 3511 */ 3512 static inline unsigned int netif_attrmask_next(int n, const unsigned long *srcp, 3513 unsigned int nr_bits) 3514 { 3515 /* -1 is a legal arg here. */ 3516 if (n != -1) 3517 cpu_max_bits_warn(n, nr_bits); 3518 3519 if (srcp) 3520 return find_next_bit(srcp, nr_bits, n + 1); 3521 3522 return n + 1; 3523 } 3524 3525 /** 3526 * netif_attrmask_next_and - get the next CPU/Rx queue in \*src1p & \*src2p 3527 * @n: CPU/Rx queue index 3528 * @src1p: the first CPUs/Rx queues mask pointer 3529 * @src2p: the second CPUs/Rx queues mask pointer 3530 * @nr_bits: number of bits in the bitmask 3531 * 3532 * Returns >= nr_bits if no further CPUs/Rx queues set in both. 3533 */ 3534 static inline int netif_attrmask_next_and(int n, const unsigned long *src1p, 3535 const unsigned long *src2p, 3536 unsigned int nr_bits) 3537 { 3538 /* -1 is a legal arg here. */ 3539 if (n != -1) 3540 cpu_max_bits_warn(n, nr_bits); 3541 3542 if (src1p && src2p) 3543 return find_next_and_bit(src1p, src2p, nr_bits, n + 1); 3544 else if (src1p) 3545 return find_next_bit(src1p, nr_bits, n + 1); 3546 else if (src2p) 3547 return find_next_bit(src2p, nr_bits, n + 1); 3548 3549 return n + 1; 3550 } 3551 #else 3552 static inline int netif_set_xps_queue(struct net_device *dev, 3553 const struct cpumask *mask, 3554 u16 index) 3555 { 3556 return 0; 3557 } 3558 3559 static inline int __netif_set_xps_queue(struct net_device *dev, 3560 const unsigned long *mask, 3561 u16 index, enum xps_map_type type) 3562 { 3563 return 0; 3564 } 3565 #endif 3566 3567 /** 3568 * netif_is_multiqueue - test if device has multiple transmit queues 3569 * @dev: network device 3570 * 3571 * Check if device has multiple transmit queues 3572 */ 3573 static inline bool netif_is_multiqueue(const struct net_device *dev) 3574 { 3575 return dev->num_tx_queues > 1; 3576 } 3577 3578 int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq); 3579 3580 #ifdef CONFIG_SYSFS 3581 int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq); 3582 #else 3583 static inline int netif_set_real_num_rx_queues(struct net_device *dev, 3584 unsigned int rxqs) 3585 { 3586 dev->real_num_rx_queues = rxqs; 3587 return 0; 3588 } 3589 #endif 3590 int netif_set_real_num_queues(struct net_device *dev, 3591 unsigned int txq, unsigned int rxq); 3592 3593 static inline struct netdev_rx_queue * 3594 __netif_get_rx_queue(struct net_device *dev, unsigned int rxq) 3595 { 3596 return dev->_rx + rxq; 3597 } 3598 3599 #ifdef CONFIG_SYSFS 3600 static inline unsigned int get_netdev_rx_queue_index( 3601 struct netdev_rx_queue *queue) 3602 { 3603 struct net_device *dev = queue->dev; 3604 int index = queue - dev->_rx; 3605 3606 BUG_ON(index >= dev->num_rx_queues); 3607 return index; 3608 } 3609 #endif 3610 3611 #define DEFAULT_MAX_NUM_RSS_QUEUES (8) 3612 int netif_get_num_default_rss_queues(void); 3613 3614 enum skb_free_reason { 3615 SKB_REASON_CONSUMED, 3616 SKB_REASON_DROPPED, 3617 }; 3618 3619 void __dev_kfree_skb_irq(struct sk_buff *skb, enum skb_free_reason reason); 3620 void __dev_kfree_skb_any(struct sk_buff *skb, enum skb_free_reason reason); 3621 3622 /* 3623 * It is not allowed to call kfree_skb() or consume_skb() from hardware 3624 * interrupt context or with hardware interrupts being disabled. 3625 * (in_hardirq() || irqs_disabled()) 3626 * 3627 * We provide four helpers that can be used in following contexts : 3628 * 3629 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context, 3630 * replacing kfree_skb(skb) 3631 * 3632 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context. 3633 * Typically used in place of consume_skb(skb) in TX completion path 3634 * 3635 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context, 3636 * replacing kfree_skb(skb) 3637 * 3638 * dev_consume_skb_any(skb) when caller doesn't know its current irq context, 3639 * and consumed a packet. Used in place of consume_skb(skb) 3640 */ 3641 static inline void dev_kfree_skb_irq(struct sk_buff *skb) 3642 { 3643 __dev_kfree_skb_irq(skb, SKB_REASON_DROPPED); 3644 } 3645 3646 static inline void dev_consume_skb_irq(struct sk_buff *skb) 3647 { 3648 __dev_kfree_skb_irq(skb, SKB_REASON_CONSUMED); 3649 } 3650 3651 static inline void dev_kfree_skb_any(struct sk_buff *skb) 3652 { 3653 __dev_kfree_skb_any(skb, SKB_REASON_DROPPED); 3654 } 3655 3656 static inline void dev_consume_skb_any(struct sk_buff *skb) 3657 { 3658 __dev_kfree_skb_any(skb, SKB_REASON_CONSUMED); 3659 } 3660 3661 u32 bpf_prog_run_generic_xdp(struct sk_buff *skb, struct xdp_buff *xdp, 3662 struct bpf_prog *xdp_prog); 3663 void generic_xdp_tx(struct sk_buff *skb, struct bpf_prog *xdp_prog); 3664 int do_xdp_generic(struct bpf_prog *xdp_prog, struct sk_buff *skb); 3665 int netif_rx(struct sk_buff *skb); 3666 int netif_rx_ni(struct sk_buff *skb); 3667 int netif_rx_any_context(struct sk_buff *skb); 3668 int netif_receive_skb(struct sk_buff *skb); 3669 int netif_receive_skb_core(struct sk_buff *skb); 3670 void netif_receive_skb_list_internal(struct list_head *head); 3671 void netif_receive_skb_list(struct list_head *head); 3672 gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb); 3673 void napi_gro_flush(struct napi_struct *napi, bool flush_old); 3674 struct sk_buff *napi_get_frags(struct napi_struct *napi); 3675 gro_result_t napi_gro_frags(struct napi_struct *napi); 3676 struct packet_offload *gro_find_receive_by_type(__be16 type); 3677 struct packet_offload *gro_find_complete_by_type(__be16 type); 3678 3679 static inline void napi_free_frags(struct napi_struct *napi) 3680 { 3681 kfree_skb(napi->skb); 3682 napi->skb = NULL; 3683 } 3684 3685 bool netdev_is_rx_handler_busy(struct net_device *dev); 3686 int netdev_rx_handler_register(struct net_device *dev, 3687 rx_handler_func_t *rx_handler, 3688 void *rx_handler_data); 3689 void netdev_rx_handler_unregister(struct net_device *dev); 3690 3691 bool dev_valid_name(const char *name); 3692 static inline bool is_socket_ioctl_cmd(unsigned int cmd) 3693 { 3694 return _IOC_TYPE(cmd) == SOCK_IOC_TYPE; 3695 } 3696 int get_user_ifreq(struct ifreq *ifr, void __user **ifrdata, void __user *arg); 3697 int put_user_ifreq(struct ifreq *ifr, void __user *arg); 3698 int dev_ioctl(struct net *net, unsigned int cmd, struct ifreq *ifr, 3699 void __user *data, bool *need_copyout); 3700 int dev_ifconf(struct net *net, struct ifconf __user *ifc); 3701 int dev_ethtool(struct net *net, struct ifreq *ifr, void __user *userdata); 3702 unsigned int dev_get_flags(const struct net_device *); 3703 int __dev_change_flags(struct net_device *dev, unsigned int flags, 3704 struct netlink_ext_ack *extack); 3705 int dev_change_flags(struct net_device *dev, unsigned int flags, 3706 struct netlink_ext_ack *extack); 3707 void __dev_notify_flags(struct net_device *, unsigned int old_flags, 3708 unsigned int gchanges); 3709 int dev_change_name(struct net_device *, const char *); 3710 int dev_set_alias(struct net_device *, const char *, size_t); 3711 int dev_get_alias(const struct net_device *, char *, size_t); 3712 int __dev_change_net_namespace(struct net_device *dev, struct net *net, 3713 const char *pat, int new_ifindex); 3714 static inline 3715 int dev_change_net_namespace(struct net_device *dev, struct net *net, 3716 const char *pat) 3717 { 3718 return __dev_change_net_namespace(dev, net, pat, 0); 3719 } 3720 int __dev_set_mtu(struct net_device *, int); 3721 int dev_validate_mtu(struct net_device *dev, int mtu, 3722 struct netlink_ext_ack *extack); 3723 int dev_set_mtu_ext(struct net_device *dev, int mtu, 3724 struct netlink_ext_ack *extack); 3725 int dev_set_mtu(struct net_device *, int); 3726 int dev_change_tx_queue_len(struct net_device *, unsigned long); 3727 void dev_set_group(struct net_device *, int); 3728 int dev_pre_changeaddr_notify(struct net_device *dev, const char *addr, 3729 struct netlink_ext_ack *extack); 3730 int dev_set_mac_address(struct net_device *dev, struct sockaddr *sa, 3731 struct netlink_ext_ack *extack); 3732 int dev_set_mac_address_user(struct net_device *dev, struct sockaddr *sa, 3733 struct netlink_ext_ack *extack); 3734 int dev_get_mac_address(struct sockaddr *sa, struct net *net, char *dev_name); 3735 int dev_change_carrier(struct net_device *, bool new_carrier); 3736 int dev_get_phys_port_id(struct net_device *dev, 3737 struct netdev_phys_item_id *ppid); 3738 int dev_get_phys_port_name(struct net_device *dev, 3739 char *name, size_t len); 3740 int dev_get_port_parent_id(struct net_device *dev, 3741 struct netdev_phys_item_id *ppid, bool recurse); 3742 bool netdev_port_same_parent_id(struct net_device *a, struct net_device *b); 3743 int dev_change_proto_down(struct net_device *dev, bool proto_down); 3744 void dev_change_proto_down_reason(struct net_device *dev, unsigned long mask, 3745 u32 value); 3746 struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev, bool *again); 3747 struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev, 3748 struct netdev_queue *txq, int *ret); 3749 3750 typedef int (*bpf_op_t)(struct net_device *dev, struct netdev_bpf *bpf); 3751 int dev_change_xdp_fd(struct net_device *dev, struct netlink_ext_ack *extack, 3752 int fd, int expected_fd, u32 flags); 3753 int bpf_xdp_link_attach(const union bpf_attr *attr, struct bpf_prog *prog); 3754 u8 dev_xdp_prog_count(struct net_device *dev); 3755 u32 dev_xdp_prog_id(struct net_device *dev, enum bpf_xdp_mode mode); 3756 3757 int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb); 3758 int dev_forward_skb(struct net_device *dev, struct sk_buff *skb); 3759 int dev_forward_skb_nomtu(struct net_device *dev, struct sk_buff *skb); 3760 bool is_skb_forwardable(const struct net_device *dev, 3761 const struct sk_buff *skb); 3762 3763 static __always_inline bool __is_skb_forwardable(const struct net_device *dev, 3764 const struct sk_buff *skb, 3765 const bool check_mtu) 3766 { 3767 const u32 vlan_hdr_len = 4; /* VLAN_HLEN */ 3768 unsigned int len; 3769 3770 if (!(dev->flags & IFF_UP)) 3771 return false; 3772 3773 if (!check_mtu) 3774 return true; 3775 3776 len = dev->mtu + dev->hard_header_len + vlan_hdr_len; 3777 if (skb->len <= len) 3778 return true; 3779 3780 /* if TSO is enabled, we don't care about the length as the packet 3781 * could be forwarded without being segmented before 3782 */ 3783 if (skb_is_gso(skb)) 3784 return true; 3785 3786 return false; 3787 } 3788 3789 static __always_inline int ____dev_forward_skb(struct net_device *dev, 3790 struct sk_buff *skb, 3791 const bool check_mtu) 3792 { 3793 if (skb_orphan_frags(skb, GFP_ATOMIC) || 3794 unlikely(!__is_skb_forwardable(dev, skb, check_mtu))) { 3795 atomic_long_inc(&dev->rx_dropped); 3796 kfree_skb(skb); 3797 return NET_RX_DROP; 3798 } 3799 3800 skb_scrub_packet(skb, !net_eq(dev_net(dev), dev_net(skb->dev))); 3801 skb->priority = 0; 3802 return 0; 3803 } 3804 3805 bool dev_nit_active(struct net_device *dev); 3806 void dev_queue_xmit_nit(struct sk_buff *skb, struct net_device *dev); 3807 3808 extern int netdev_budget; 3809 extern unsigned int netdev_budget_usecs; 3810 3811 /* Called by rtnetlink.c:rtnl_unlock() */ 3812 void netdev_run_todo(void); 3813 3814 /** 3815 * dev_put - release reference to device 3816 * @dev: network device 3817 * 3818 * Release reference to device to allow it to be freed. 3819 * Try using dev_put_track() instead. 3820 */ 3821 static inline void dev_put(struct net_device *dev) 3822 { 3823 if (dev) { 3824 #ifdef CONFIG_PCPU_DEV_REFCNT 3825 this_cpu_dec(*dev->pcpu_refcnt); 3826 #else 3827 refcount_dec(&dev->dev_refcnt); 3828 #endif 3829 } 3830 } 3831 3832 /** 3833 * dev_hold - get reference to device 3834 * @dev: network device 3835 * 3836 * Hold reference to device to keep it from being freed. 3837 * Try using dev_hold_track() instead. 3838 */ 3839 static inline void dev_hold(struct net_device *dev) 3840 { 3841 if (dev) { 3842 #ifdef CONFIG_PCPU_DEV_REFCNT 3843 this_cpu_inc(*dev->pcpu_refcnt); 3844 #else 3845 refcount_inc(&dev->dev_refcnt); 3846 #endif 3847 } 3848 } 3849 3850 static inline void netdev_tracker_alloc(struct net_device *dev, 3851 netdevice_tracker *tracker, gfp_t gfp) 3852 { 3853 #ifdef CONFIG_NET_DEV_REFCNT_TRACKER 3854 ref_tracker_alloc(&dev->refcnt_tracker, tracker, gfp); 3855 #endif 3856 } 3857 3858 static inline void netdev_tracker_free(struct net_device *dev, 3859 netdevice_tracker *tracker) 3860 { 3861 #ifdef CONFIG_NET_DEV_REFCNT_TRACKER 3862 ref_tracker_free(&dev->refcnt_tracker, tracker); 3863 #endif 3864 } 3865 3866 static inline void dev_hold_track(struct net_device *dev, 3867 netdevice_tracker *tracker, gfp_t gfp) 3868 { 3869 if (dev) { 3870 dev_hold(dev); 3871 netdev_tracker_alloc(dev, tracker, gfp); 3872 } 3873 } 3874 3875 static inline void dev_put_track(struct net_device *dev, 3876 netdevice_tracker *tracker) 3877 { 3878 if (dev) { 3879 netdev_tracker_free(dev, tracker); 3880 dev_put(dev); 3881 } 3882 } 3883 3884 static inline void dev_replace_track(struct net_device *odev, 3885 struct net_device *ndev, 3886 netdevice_tracker *tracker, 3887 gfp_t gfp) 3888 { 3889 if (odev) 3890 netdev_tracker_free(odev, tracker); 3891 3892 dev_hold(ndev); 3893 dev_put(odev); 3894 3895 if (ndev) 3896 netdev_tracker_alloc(ndev, tracker, gfp); 3897 } 3898 3899 /* Carrier loss detection, dial on demand. The functions netif_carrier_on 3900 * and _off may be called from IRQ context, but it is caller 3901 * who is responsible for serialization of these calls. 3902 * 3903 * The name carrier is inappropriate, these functions should really be 3904 * called netif_lowerlayer_*() because they represent the state of any 3905 * kind of lower layer not just hardware media. 3906 */ 3907 3908 void linkwatch_init_dev(struct net_device *dev); 3909 void linkwatch_fire_event(struct net_device *dev); 3910 void linkwatch_forget_dev(struct net_device *dev); 3911 3912 /** 3913 * netif_carrier_ok - test if carrier present 3914 * @dev: network device 3915 * 3916 * Check if carrier is present on device 3917 */ 3918 static inline bool netif_carrier_ok(const struct net_device *dev) 3919 { 3920 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state); 3921 } 3922 3923 unsigned long dev_trans_start(struct net_device *dev); 3924 3925 void __netdev_watchdog_up(struct net_device *dev); 3926 3927 void netif_carrier_on(struct net_device *dev); 3928 void netif_carrier_off(struct net_device *dev); 3929 void netif_carrier_event(struct net_device *dev); 3930 3931 /** 3932 * netif_dormant_on - mark device as dormant. 3933 * @dev: network device 3934 * 3935 * Mark device as dormant (as per RFC2863). 3936 * 3937 * The dormant state indicates that the relevant interface is not 3938 * actually in a condition to pass packets (i.e., it is not 'up') but is 3939 * in a "pending" state, waiting for some external event. For "on- 3940 * demand" interfaces, this new state identifies the situation where the 3941 * interface is waiting for events to place it in the up state. 3942 */ 3943 static inline void netif_dormant_on(struct net_device *dev) 3944 { 3945 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state)) 3946 linkwatch_fire_event(dev); 3947 } 3948 3949 /** 3950 * netif_dormant_off - set device as not dormant. 3951 * @dev: network device 3952 * 3953 * Device is not in dormant state. 3954 */ 3955 static inline void netif_dormant_off(struct net_device *dev) 3956 { 3957 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state)) 3958 linkwatch_fire_event(dev); 3959 } 3960 3961 /** 3962 * netif_dormant - test if device is dormant 3963 * @dev: network device 3964 * 3965 * Check if device is dormant. 3966 */ 3967 static inline bool netif_dormant(const struct net_device *dev) 3968 { 3969 return test_bit(__LINK_STATE_DORMANT, &dev->state); 3970 } 3971 3972 3973 /** 3974 * netif_testing_on - mark device as under test. 3975 * @dev: network device 3976 * 3977 * Mark device as under test (as per RFC2863). 3978 * 3979 * The testing state indicates that some test(s) must be performed on 3980 * the interface. After completion, of the test, the interface state 3981 * will change to up, dormant, or down, as appropriate. 3982 */ 3983 static inline void netif_testing_on(struct net_device *dev) 3984 { 3985 if (!test_and_set_bit(__LINK_STATE_TESTING, &dev->state)) 3986 linkwatch_fire_event(dev); 3987 } 3988 3989 /** 3990 * netif_testing_off - set device as not under test. 3991 * @dev: network device 3992 * 3993 * Device is not in testing state. 3994 */ 3995 static inline void netif_testing_off(struct net_device *dev) 3996 { 3997 if (test_and_clear_bit(__LINK_STATE_TESTING, &dev->state)) 3998 linkwatch_fire_event(dev); 3999 } 4000 4001 /** 4002 * netif_testing - test if device is under test 4003 * @dev: network device 4004 * 4005 * Check if device is under test 4006 */ 4007 static inline bool netif_testing(const struct net_device *dev) 4008 { 4009 return test_bit(__LINK_STATE_TESTING, &dev->state); 4010 } 4011 4012 4013 /** 4014 * netif_oper_up - test if device is operational 4015 * @dev: network device 4016 * 4017 * Check if carrier is operational 4018 */ 4019 static inline bool netif_oper_up(const struct net_device *dev) 4020 { 4021 return (dev->operstate == IF_OPER_UP || 4022 dev->operstate == IF_OPER_UNKNOWN /* backward compat */); 4023 } 4024 4025 /** 4026 * netif_device_present - is device available or removed 4027 * @dev: network device 4028 * 4029 * Check if device has not been removed from system. 4030 */ 4031 static inline bool netif_device_present(const struct net_device *dev) 4032 { 4033 return test_bit(__LINK_STATE_PRESENT, &dev->state); 4034 } 4035 4036 void netif_device_detach(struct net_device *dev); 4037 4038 void netif_device_attach(struct net_device *dev); 4039 4040 /* 4041 * Network interface message level settings 4042 */ 4043 4044 enum { 4045 NETIF_MSG_DRV_BIT, 4046 NETIF_MSG_PROBE_BIT, 4047 NETIF_MSG_LINK_BIT, 4048 NETIF_MSG_TIMER_BIT, 4049 NETIF_MSG_IFDOWN_BIT, 4050 NETIF_MSG_IFUP_BIT, 4051 NETIF_MSG_RX_ERR_BIT, 4052 NETIF_MSG_TX_ERR_BIT, 4053 NETIF_MSG_TX_QUEUED_BIT, 4054 NETIF_MSG_INTR_BIT, 4055 NETIF_MSG_TX_DONE_BIT, 4056 NETIF_MSG_RX_STATUS_BIT, 4057 NETIF_MSG_PKTDATA_BIT, 4058 NETIF_MSG_HW_BIT, 4059 NETIF_MSG_WOL_BIT, 4060 4061 /* When you add a new bit above, update netif_msg_class_names array 4062 * in net/ethtool/common.c 4063 */ 4064 NETIF_MSG_CLASS_COUNT, 4065 }; 4066 /* Both ethtool_ops interface and internal driver implementation use u32 */ 4067 static_assert(NETIF_MSG_CLASS_COUNT <= 32); 4068 4069 #define __NETIF_MSG_BIT(bit) ((u32)1 << (bit)) 4070 #define __NETIF_MSG(name) __NETIF_MSG_BIT(NETIF_MSG_ ## name ## _BIT) 4071 4072 #define NETIF_MSG_DRV __NETIF_MSG(DRV) 4073 #define NETIF_MSG_PROBE __NETIF_MSG(PROBE) 4074 #define NETIF_MSG_LINK __NETIF_MSG(LINK) 4075 #define NETIF_MSG_TIMER __NETIF_MSG(TIMER) 4076 #define NETIF_MSG_IFDOWN __NETIF_MSG(IFDOWN) 4077 #define NETIF_MSG_IFUP __NETIF_MSG(IFUP) 4078 #define NETIF_MSG_RX_ERR __NETIF_MSG(RX_ERR) 4079 #define NETIF_MSG_TX_ERR __NETIF_MSG(TX_ERR) 4080 #define NETIF_MSG_TX_QUEUED __NETIF_MSG(TX_QUEUED) 4081 #define NETIF_MSG_INTR __NETIF_MSG(INTR) 4082 #define NETIF_MSG_TX_DONE __NETIF_MSG(TX_DONE) 4083 #define NETIF_MSG_RX_STATUS __NETIF_MSG(RX_STATUS) 4084 #define NETIF_MSG_PKTDATA __NETIF_MSG(PKTDATA) 4085 #define NETIF_MSG_HW __NETIF_MSG(HW) 4086 #define NETIF_MSG_WOL __NETIF_MSG(WOL) 4087 4088 #define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV) 4089 #define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE) 4090 #define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK) 4091 #define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER) 4092 #define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN) 4093 #define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP) 4094 #define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR) 4095 #define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR) 4096 #define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED) 4097 #define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR) 4098 #define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE) 4099 #define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS) 4100 #define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA) 4101 #define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW) 4102 #define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL) 4103 4104 static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits) 4105 { 4106 /* use default */ 4107 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8)) 4108 return default_msg_enable_bits; 4109 if (debug_value == 0) /* no output */ 4110 return 0; 4111 /* set low N bits */ 4112 return (1U << debug_value) - 1; 4113 } 4114 4115 static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu) 4116 { 4117 spin_lock(&txq->_xmit_lock); 4118 /* Pairs with READ_ONCE() in __dev_queue_xmit() */ 4119 WRITE_ONCE(txq->xmit_lock_owner, cpu); 4120 } 4121 4122 static inline bool __netif_tx_acquire(struct netdev_queue *txq) 4123 { 4124 __acquire(&txq->_xmit_lock); 4125 return true; 4126 } 4127 4128 static inline void __netif_tx_release(struct netdev_queue *txq) 4129 { 4130 __release(&txq->_xmit_lock); 4131 } 4132 4133 static inline void __netif_tx_lock_bh(struct netdev_queue *txq) 4134 { 4135 spin_lock_bh(&txq->_xmit_lock); 4136 /* Pairs with READ_ONCE() in __dev_queue_xmit() */ 4137 WRITE_ONCE(txq->xmit_lock_owner, smp_processor_id()); 4138 } 4139 4140 static inline bool __netif_tx_trylock(struct netdev_queue *txq) 4141 { 4142 bool ok = spin_trylock(&txq->_xmit_lock); 4143 4144 if (likely(ok)) { 4145 /* Pairs with READ_ONCE() in __dev_queue_xmit() */ 4146 WRITE_ONCE(txq->xmit_lock_owner, smp_processor_id()); 4147 } 4148 return ok; 4149 } 4150 4151 static inline void __netif_tx_unlock(struct netdev_queue *txq) 4152 { 4153 /* Pairs with READ_ONCE() in __dev_queue_xmit() */ 4154 WRITE_ONCE(txq->xmit_lock_owner, -1); 4155 spin_unlock(&txq->_xmit_lock); 4156 } 4157 4158 static inline void __netif_tx_unlock_bh(struct netdev_queue *txq) 4159 { 4160 /* Pairs with READ_ONCE() in __dev_queue_xmit() */ 4161 WRITE_ONCE(txq->xmit_lock_owner, -1); 4162 spin_unlock_bh(&txq->_xmit_lock); 4163 } 4164 4165 /* 4166 * txq->trans_start can be read locklessly from dev_watchdog() 4167 */ 4168 static inline void txq_trans_update(struct netdev_queue *txq) 4169 { 4170 if (txq->xmit_lock_owner != -1) 4171 WRITE_ONCE(txq->trans_start, jiffies); 4172 } 4173 4174 static inline void txq_trans_cond_update(struct netdev_queue *txq) 4175 { 4176 unsigned long now = jiffies; 4177 4178 if (READ_ONCE(txq->trans_start) != now) 4179 WRITE_ONCE(txq->trans_start, now); 4180 } 4181 4182 /* legacy drivers only, netdev_start_xmit() sets txq->trans_start */ 4183 static inline void netif_trans_update(struct net_device *dev) 4184 { 4185 struct netdev_queue *txq = netdev_get_tx_queue(dev, 0); 4186 4187 txq_trans_cond_update(txq); 4188 } 4189 4190 /** 4191 * netif_tx_lock - grab network device transmit lock 4192 * @dev: network device 4193 * 4194 * Get network device transmit lock 4195 */ 4196 void netif_tx_lock(struct net_device *dev); 4197 4198 static inline void netif_tx_lock_bh(struct net_device *dev) 4199 { 4200 local_bh_disable(); 4201 netif_tx_lock(dev); 4202 } 4203 4204 void netif_tx_unlock(struct net_device *dev); 4205 4206 static inline void netif_tx_unlock_bh(struct net_device *dev) 4207 { 4208 netif_tx_unlock(dev); 4209 local_bh_enable(); 4210 } 4211 4212 #define HARD_TX_LOCK(dev, txq, cpu) { \ 4213 if ((dev->features & NETIF_F_LLTX) == 0) { \ 4214 __netif_tx_lock(txq, cpu); \ 4215 } else { \ 4216 __netif_tx_acquire(txq); \ 4217 } \ 4218 } 4219 4220 #define HARD_TX_TRYLOCK(dev, txq) \ 4221 (((dev->features & NETIF_F_LLTX) == 0) ? \ 4222 __netif_tx_trylock(txq) : \ 4223 __netif_tx_acquire(txq)) 4224 4225 #define HARD_TX_UNLOCK(dev, txq) { \ 4226 if ((dev->features & NETIF_F_LLTX) == 0) { \ 4227 __netif_tx_unlock(txq); \ 4228 } else { \ 4229 __netif_tx_release(txq); \ 4230 } \ 4231 } 4232 4233 static inline void netif_tx_disable(struct net_device *dev) 4234 { 4235 unsigned int i; 4236 int cpu; 4237 4238 local_bh_disable(); 4239 cpu = smp_processor_id(); 4240 spin_lock(&dev->tx_global_lock); 4241 for (i = 0; i < dev->num_tx_queues; i++) { 4242 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 4243 4244 __netif_tx_lock(txq, cpu); 4245 netif_tx_stop_queue(txq); 4246 __netif_tx_unlock(txq); 4247 } 4248 spin_unlock(&dev->tx_global_lock); 4249 local_bh_enable(); 4250 } 4251 4252 static inline void netif_addr_lock(struct net_device *dev) 4253 { 4254 unsigned char nest_level = 0; 4255 4256 #ifdef CONFIG_LOCKDEP 4257 nest_level = dev->nested_level; 4258 #endif 4259 spin_lock_nested(&dev->addr_list_lock, nest_level); 4260 } 4261 4262 static inline void netif_addr_lock_bh(struct net_device *dev) 4263 { 4264 unsigned char nest_level = 0; 4265 4266 #ifdef CONFIG_LOCKDEP 4267 nest_level = dev->nested_level; 4268 #endif 4269 local_bh_disable(); 4270 spin_lock_nested(&dev->addr_list_lock, nest_level); 4271 } 4272 4273 static inline void netif_addr_unlock(struct net_device *dev) 4274 { 4275 spin_unlock(&dev->addr_list_lock); 4276 } 4277 4278 static inline void netif_addr_unlock_bh(struct net_device *dev) 4279 { 4280 spin_unlock_bh(&dev->addr_list_lock); 4281 } 4282 4283 /* 4284 * dev_addrs walker. Should be used only for read access. Call with 4285 * rcu_read_lock held. 4286 */ 4287 #define for_each_dev_addr(dev, ha) \ 4288 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list) 4289 4290 /* These functions live elsewhere (drivers/net/net_init.c, but related) */ 4291 4292 void ether_setup(struct net_device *dev); 4293 4294 /* Support for loadable net-drivers */ 4295 struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name, 4296 unsigned char name_assign_type, 4297 void (*setup)(struct net_device *), 4298 unsigned int txqs, unsigned int rxqs); 4299 #define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \ 4300 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1) 4301 4302 #define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \ 4303 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \ 4304 count) 4305 4306 int register_netdev(struct net_device *dev); 4307 void unregister_netdev(struct net_device *dev); 4308 4309 int devm_register_netdev(struct device *dev, struct net_device *ndev); 4310 4311 /* General hardware address lists handling functions */ 4312 int __hw_addr_sync(struct netdev_hw_addr_list *to_list, 4313 struct netdev_hw_addr_list *from_list, int addr_len); 4314 void __hw_addr_unsync(struct netdev_hw_addr_list *to_list, 4315 struct netdev_hw_addr_list *from_list, int addr_len); 4316 int __hw_addr_sync_dev(struct netdev_hw_addr_list *list, 4317 struct net_device *dev, 4318 int (*sync)(struct net_device *, const unsigned char *), 4319 int (*unsync)(struct net_device *, 4320 const unsigned char *)); 4321 int __hw_addr_ref_sync_dev(struct netdev_hw_addr_list *list, 4322 struct net_device *dev, 4323 int (*sync)(struct net_device *, 4324 const unsigned char *, int), 4325 int (*unsync)(struct net_device *, 4326 const unsigned char *, int)); 4327 void __hw_addr_ref_unsync_dev(struct netdev_hw_addr_list *list, 4328 struct net_device *dev, 4329 int (*unsync)(struct net_device *, 4330 const unsigned char *, int)); 4331 void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list, 4332 struct net_device *dev, 4333 int (*unsync)(struct net_device *, 4334 const unsigned char *)); 4335 void __hw_addr_init(struct netdev_hw_addr_list *list); 4336 4337 /* Functions used for device addresses handling */ 4338 void dev_addr_mod(struct net_device *dev, unsigned int offset, 4339 const void *addr, size_t len); 4340 4341 static inline void 4342 __dev_addr_set(struct net_device *dev, const void *addr, size_t len) 4343 { 4344 dev_addr_mod(dev, 0, addr, len); 4345 } 4346 4347 static inline void dev_addr_set(struct net_device *dev, const u8 *addr) 4348 { 4349 __dev_addr_set(dev, addr, dev->addr_len); 4350 } 4351 4352 int dev_addr_add(struct net_device *dev, const unsigned char *addr, 4353 unsigned char addr_type); 4354 int dev_addr_del(struct net_device *dev, const unsigned char *addr, 4355 unsigned char addr_type); 4356 void dev_addr_flush(struct net_device *dev); 4357 int dev_addr_init(struct net_device *dev); 4358 void dev_addr_check(struct net_device *dev); 4359 4360 /* Functions used for unicast addresses handling */ 4361 int dev_uc_add(struct net_device *dev, const unsigned char *addr); 4362 int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr); 4363 int dev_uc_del(struct net_device *dev, const unsigned char *addr); 4364 int dev_uc_sync(struct net_device *to, struct net_device *from); 4365 int dev_uc_sync_multiple(struct net_device *to, struct net_device *from); 4366 void dev_uc_unsync(struct net_device *to, struct net_device *from); 4367 void dev_uc_flush(struct net_device *dev); 4368 void dev_uc_init(struct net_device *dev); 4369 4370 /** 4371 * __dev_uc_sync - Synchonize device's unicast list 4372 * @dev: device to sync 4373 * @sync: function to call if address should be added 4374 * @unsync: function to call if address should be removed 4375 * 4376 * Add newly added addresses to the interface, and release 4377 * addresses that have been deleted. 4378 */ 4379 static inline int __dev_uc_sync(struct net_device *dev, 4380 int (*sync)(struct net_device *, 4381 const unsigned char *), 4382 int (*unsync)(struct net_device *, 4383 const unsigned char *)) 4384 { 4385 return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync); 4386 } 4387 4388 /** 4389 * __dev_uc_unsync - Remove synchronized addresses from device 4390 * @dev: device to sync 4391 * @unsync: function to call if address should be removed 4392 * 4393 * Remove all addresses that were added to the device by dev_uc_sync(). 4394 */ 4395 static inline void __dev_uc_unsync(struct net_device *dev, 4396 int (*unsync)(struct net_device *, 4397 const unsigned char *)) 4398 { 4399 __hw_addr_unsync_dev(&dev->uc, dev, unsync); 4400 } 4401 4402 /* Functions used for multicast addresses handling */ 4403 int dev_mc_add(struct net_device *dev, const unsigned char *addr); 4404 int dev_mc_add_global(struct net_device *dev, const unsigned char *addr); 4405 int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr); 4406 int dev_mc_del(struct net_device *dev, const unsigned char *addr); 4407 int dev_mc_del_global(struct net_device *dev, const unsigned char *addr); 4408 int dev_mc_sync(struct net_device *to, struct net_device *from); 4409 int dev_mc_sync_multiple(struct net_device *to, struct net_device *from); 4410 void dev_mc_unsync(struct net_device *to, struct net_device *from); 4411 void dev_mc_flush(struct net_device *dev); 4412 void dev_mc_init(struct net_device *dev); 4413 4414 /** 4415 * __dev_mc_sync - Synchonize device's multicast 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_mc_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->mc, dev, sync, unsync); 4430 } 4431 4432 /** 4433 * __dev_mc_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_mc_sync(). 4438 */ 4439 static inline void __dev_mc_unsync(struct net_device *dev, 4440 int (*unsync)(struct net_device *, 4441 const unsigned char *)) 4442 { 4443 __hw_addr_unsync_dev(&dev->mc, dev, unsync); 4444 } 4445 4446 /* Functions used for secondary unicast and multicast support */ 4447 void dev_set_rx_mode(struct net_device *dev); 4448 void __dev_set_rx_mode(struct net_device *dev); 4449 int dev_set_promiscuity(struct net_device *dev, int inc); 4450 int dev_set_allmulti(struct net_device *dev, int inc); 4451 void netdev_state_change(struct net_device *dev); 4452 void __netdev_notify_peers(struct net_device *dev); 4453 void netdev_notify_peers(struct net_device *dev); 4454 void netdev_features_change(struct net_device *dev); 4455 /* Load a device via the kmod */ 4456 void dev_load(struct net *net, const char *name); 4457 struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev, 4458 struct rtnl_link_stats64 *storage); 4459 void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64, 4460 const struct net_device_stats *netdev_stats); 4461 void dev_fetch_sw_netstats(struct rtnl_link_stats64 *s, 4462 const struct pcpu_sw_netstats __percpu *netstats); 4463 void dev_get_tstats64(struct net_device *dev, struct rtnl_link_stats64 *s); 4464 4465 extern int netdev_max_backlog; 4466 extern int netdev_tstamp_prequeue; 4467 extern int netdev_unregister_timeout_secs; 4468 extern int weight_p; 4469 extern int dev_weight_rx_bias; 4470 extern int dev_weight_tx_bias; 4471 extern int dev_rx_weight; 4472 extern int dev_tx_weight; 4473 extern int gro_normal_batch; 4474 4475 enum { 4476 NESTED_SYNC_IMM_BIT, 4477 NESTED_SYNC_TODO_BIT, 4478 }; 4479 4480 #define __NESTED_SYNC_BIT(bit) ((u32)1 << (bit)) 4481 #define __NESTED_SYNC(name) __NESTED_SYNC_BIT(NESTED_SYNC_ ## name ## _BIT) 4482 4483 #define NESTED_SYNC_IMM __NESTED_SYNC(IMM) 4484 #define NESTED_SYNC_TODO __NESTED_SYNC(TODO) 4485 4486 struct netdev_nested_priv { 4487 unsigned char flags; 4488 void *data; 4489 }; 4490 4491 bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev); 4492 struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev, 4493 struct list_head **iter); 4494 4495 #ifdef CONFIG_LOCKDEP 4496 static LIST_HEAD(net_unlink_list); 4497 4498 static inline void net_unlink_todo(struct net_device *dev) 4499 { 4500 if (list_empty(&dev->unlink_list)) 4501 list_add_tail(&dev->unlink_list, &net_unlink_list); 4502 } 4503 #endif 4504 4505 /* iterate through upper list, must be called under RCU read lock */ 4506 #define netdev_for_each_upper_dev_rcu(dev, updev, iter) \ 4507 for (iter = &(dev)->adj_list.upper, \ 4508 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \ 4509 updev; \ 4510 updev = netdev_upper_get_next_dev_rcu(dev, &(iter))) 4511 4512 int netdev_walk_all_upper_dev_rcu(struct net_device *dev, 4513 int (*fn)(struct net_device *upper_dev, 4514 struct netdev_nested_priv *priv), 4515 struct netdev_nested_priv *priv); 4516 4517 bool netdev_has_upper_dev_all_rcu(struct net_device *dev, 4518 struct net_device *upper_dev); 4519 4520 bool netdev_has_any_upper_dev(struct net_device *dev); 4521 4522 void *netdev_lower_get_next_private(struct net_device *dev, 4523 struct list_head **iter); 4524 void *netdev_lower_get_next_private_rcu(struct net_device *dev, 4525 struct list_head **iter); 4526 4527 #define netdev_for_each_lower_private(dev, priv, iter) \ 4528 for (iter = (dev)->adj_list.lower.next, \ 4529 priv = netdev_lower_get_next_private(dev, &(iter)); \ 4530 priv; \ 4531 priv = netdev_lower_get_next_private(dev, &(iter))) 4532 4533 #define netdev_for_each_lower_private_rcu(dev, priv, iter) \ 4534 for (iter = &(dev)->adj_list.lower, \ 4535 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \ 4536 priv; \ 4537 priv = netdev_lower_get_next_private_rcu(dev, &(iter))) 4538 4539 void *netdev_lower_get_next(struct net_device *dev, 4540 struct list_head **iter); 4541 4542 #define netdev_for_each_lower_dev(dev, ldev, iter) \ 4543 for (iter = (dev)->adj_list.lower.next, \ 4544 ldev = netdev_lower_get_next(dev, &(iter)); \ 4545 ldev; \ 4546 ldev = netdev_lower_get_next(dev, &(iter))) 4547 4548 struct net_device *netdev_next_lower_dev_rcu(struct net_device *dev, 4549 struct list_head **iter); 4550 int netdev_walk_all_lower_dev(struct net_device *dev, 4551 int (*fn)(struct net_device *lower_dev, 4552 struct netdev_nested_priv *priv), 4553 struct netdev_nested_priv *priv); 4554 int netdev_walk_all_lower_dev_rcu(struct net_device *dev, 4555 int (*fn)(struct net_device *lower_dev, 4556 struct netdev_nested_priv *priv), 4557 struct netdev_nested_priv *priv); 4558 4559 void *netdev_adjacent_get_private(struct list_head *adj_list); 4560 void *netdev_lower_get_first_private_rcu(struct net_device *dev); 4561 struct net_device *netdev_master_upper_dev_get(struct net_device *dev); 4562 struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev); 4563 int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev, 4564 struct netlink_ext_ack *extack); 4565 int netdev_master_upper_dev_link(struct net_device *dev, 4566 struct net_device *upper_dev, 4567 void *upper_priv, void *upper_info, 4568 struct netlink_ext_ack *extack); 4569 void netdev_upper_dev_unlink(struct net_device *dev, 4570 struct net_device *upper_dev); 4571 int netdev_adjacent_change_prepare(struct net_device *old_dev, 4572 struct net_device *new_dev, 4573 struct net_device *dev, 4574 struct netlink_ext_ack *extack); 4575 void netdev_adjacent_change_commit(struct net_device *old_dev, 4576 struct net_device *new_dev, 4577 struct net_device *dev); 4578 void netdev_adjacent_change_abort(struct net_device *old_dev, 4579 struct net_device *new_dev, 4580 struct net_device *dev); 4581 void netdev_adjacent_rename_links(struct net_device *dev, char *oldname); 4582 void *netdev_lower_dev_get_private(struct net_device *dev, 4583 struct net_device *lower_dev); 4584 void netdev_lower_state_changed(struct net_device *lower_dev, 4585 void *lower_state_info); 4586 4587 /* RSS keys are 40 or 52 bytes long */ 4588 #define NETDEV_RSS_KEY_LEN 52 4589 extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN] __read_mostly; 4590 void netdev_rss_key_fill(void *buffer, size_t len); 4591 4592 int skb_checksum_help(struct sk_buff *skb); 4593 int skb_crc32c_csum_help(struct sk_buff *skb); 4594 int skb_csum_hwoffload_help(struct sk_buff *skb, 4595 const netdev_features_t features); 4596 4597 struct sk_buff *__skb_gso_segment(struct sk_buff *skb, 4598 netdev_features_t features, bool tx_path); 4599 struct sk_buff *skb_mac_gso_segment(struct sk_buff *skb, 4600 netdev_features_t features); 4601 4602 struct netdev_bonding_info { 4603 ifslave slave; 4604 ifbond master; 4605 }; 4606 4607 struct netdev_notifier_bonding_info { 4608 struct netdev_notifier_info info; /* must be first */ 4609 struct netdev_bonding_info bonding_info; 4610 }; 4611 4612 void netdev_bonding_info_change(struct net_device *dev, 4613 struct netdev_bonding_info *bonding_info); 4614 4615 #if IS_ENABLED(CONFIG_ETHTOOL_NETLINK) 4616 void ethtool_notify(struct net_device *dev, unsigned int cmd, const void *data); 4617 #else 4618 static inline void ethtool_notify(struct net_device *dev, unsigned int cmd, 4619 const void *data) 4620 { 4621 } 4622 #endif 4623 4624 static inline 4625 struct sk_buff *skb_gso_segment(struct sk_buff *skb, netdev_features_t features) 4626 { 4627 return __skb_gso_segment(skb, features, true); 4628 } 4629 __be16 skb_network_protocol(struct sk_buff *skb, int *depth); 4630 4631 static inline bool can_checksum_protocol(netdev_features_t features, 4632 __be16 protocol) 4633 { 4634 if (protocol == htons(ETH_P_FCOE)) 4635 return !!(features & NETIF_F_FCOE_CRC); 4636 4637 /* Assume this is an IP checksum (not SCTP CRC) */ 4638 4639 if (features & NETIF_F_HW_CSUM) { 4640 /* Can checksum everything */ 4641 return true; 4642 } 4643 4644 switch (protocol) { 4645 case htons(ETH_P_IP): 4646 return !!(features & NETIF_F_IP_CSUM); 4647 case htons(ETH_P_IPV6): 4648 return !!(features & NETIF_F_IPV6_CSUM); 4649 default: 4650 return false; 4651 } 4652 } 4653 4654 #ifdef CONFIG_BUG 4655 void netdev_rx_csum_fault(struct net_device *dev, struct sk_buff *skb); 4656 #else 4657 static inline void netdev_rx_csum_fault(struct net_device *dev, 4658 struct sk_buff *skb) 4659 { 4660 } 4661 #endif 4662 /* rx skb timestamps */ 4663 void net_enable_timestamp(void); 4664 void net_disable_timestamp(void); 4665 4666 #ifdef CONFIG_PROC_FS 4667 int __init dev_proc_init(void); 4668 #else 4669 #define dev_proc_init() 0 4670 #endif 4671 4672 static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops, 4673 struct sk_buff *skb, struct net_device *dev, 4674 bool more) 4675 { 4676 __this_cpu_write(softnet_data.xmit.more, more); 4677 return ops->ndo_start_xmit(skb, dev); 4678 } 4679 4680 static inline bool netdev_xmit_more(void) 4681 { 4682 return __this_cpu_read(softnet_data.xmit.more); 4683 } 4684 4685 static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev, 4686 struct netdev_queue *txq, bool more) 4687 { 4688 const struct net_device_ops *ops = dev->netdev_ops; 4689 netdev_tx_t rc; 4690 4691 rc = __netdev_start_xmit(ops, skb, dev, more); 4692 if (rc == NETDEV_TX_OK) 4693 txq_trans_update(txq); 4694 4695 return rc; 4696 } 4697 4698 int netdev_class_create_file_ns(const struct class_attribute *class_attr, 4699 const void *ns); 4700 void netdev_class_remove_file_ns(const struct class_attribute *class_attr, 4701 const void *ns); 4702 4703 extern const struct kobj_ns_type_operations net_ns_type_operations; 4704 4705 const char *netdev_drivername(const struct net_device *dev); 4706 4707 void linkwatch_run_queue(void); 4708 4709 static inline netdev_features_t netdev_intersect_features(netdev_features_t f1, 4710 netdev_features_t f2) 4711 { 4712 if ((f1 ^ f2) & NETIF_F_HW_CSUM) { 4713 if (f1 & NETIF_F_HW_CSUM) 4714 f1 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM); 4715 else 4716 f2 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM); 4717 } 4718 4719 return f1 & f2; 4720 } 4721 4722 static inline netdev_features_t netdev_get_wanted_features( 4723 struct net_device *dev) 4724 { 4725 return (dev->features & ~dev->hw_features) | dev->wanted_features; 4726 } 4727 netdev_features_t netdev_increment_features(netdev_features_t all, 4728 netdev_features_t one, netdev_features_t mask); 4729 4730 /* Allow TSO being used on stacked device : 4731 * Performing the GSO segmentation before last device 4732 * is a performance improvement. 4733 */ 4734 static inline netdev_features_t netdev_add_tso_features(netdev_features_t features, 4735 netdev_features_t mask) 4736 { 4737 return netdev_increment_features(features, NETIF_F_ALL_TSO, mask); 4738 } 4739 4740 int __netdev_update_features(struct net_device *dev); 4741 void netdev_update_features(struct net_device *dev); 4742 void netdev_change_features(struct net_device *dev); 4743 4744 void netif_stacked_transfer_operstate(const struct net_device *rootdev, 4745 struct net_device *dev); 4746 4747 netdev_features_t passthru_features_check(struct sk_buff *skb, 4748 struct net_device *dev, 4749 netdev_features_t features); 4750 netdev_features_t netif_skb_features(struct sk_buff *skb); 4751 4752 static inline bool net_gso_ok(netdev_features_t features, int gso_type) 4753 { 4754 netdev_features_t feature = (netdev_features_t)gso_type << NETIF_F_GSO_SHIFT; 4755 4756 /* check flags correspondence */ 4757 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT)); 4758 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT)); 4759 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT)); 4760 BUILD_BUG_ON(SKB_GSO_TCP_FIXEDID != (NETIF_F_TSO_MANGLEID >> NETIF_F_GSO_SHIFT)); 4761 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT)); 4762 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT)); 4763 BUILD_BUG_ON(SKB_GSO_GRE != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT)); 4764 BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT)); 4765 BUILD_BUG_ON(SKB_GSO_IPXIP4 != (NETIF_F_GSO_IPXIP4 >> NETIF_F_GSO_SHIFT)); 4766 BUILD_BUG_ON(SKB_GSO_IPXIP6 != (NETIF_F_GSO_IPXIP6 >> NETIF_F_GSO_SHIFT)); 4767 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT)); 4768 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT)); 4769 BUILD_BUG_ON(SKB_GSO_PARTIAL != (NETIF_F_GSO_PARTIAL >> NETIF_F_GSO_SHIFT)); 4770 BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT)); 4771 BUILD_BUG_ON(SKB_GSO_SCTP != (NETIF_F_GSO_SCTP >> NETIF_F_GSO_SHIFT)); 4772 BUILD_BUG_ON(SKB_GSO_ESP != (NETIF_F_GSO_ESP >> NETIF_F_GSO_SHIFT)); 4773 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_GSO_UDP >> NETIF_F_GSO_SHIFT)); 4774 BUILD_BUG_ON(SKB_GSO_UDP_L4 != (NETIF_F_GSO_UDP_L4 >> NETIF_F_GSO_SHIFT)); 4775 BUILD_BUG_ON(SKB_GSO_FRAGLIST != (NETIF_F_GSO_FRAGLIST >> NETIF_F_GSO_SHIFT)); 4776 4777 return (features & feature) == feature; 4778 } 4779 4780 static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features) 4781 { 4782 return net_gso_ok(features, skb_shinfo(skb)->gso_type) && 4783 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST)); 4784 } 4785 4786 static inline bool netif_needs_gso(struct sk_buff *skb, 4787 netdev_features_t features) 4788 { 4789 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) || 4790 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) && 4791 (skb->ip_summed != CHECKSUM_UNNECESSARY))); 4792 } 4793 4794 static inline void netif_set_gso_max_size(struct net_device *dev, 4795 unsigned int size) 4796 { 4797 /* dev->gso_max_size is read locklessly from sk_setup_caps() */ 4798 WRITE_ONCE(dev->gso_max_size, size); 4799 } 4800 4801 static inline void netif_set_gso_max_segs(struct net_device *dev, 4802 unsigned int segs) 4803 { 4804 /* dev->gso_max_segs is read locklessly from sk_setup_caps() */ 4805 WRITE_ONCE(dev->gso_max_segs, segs); 4806 } 4807 4808 static inline void skb_gso_error_unwind(struct sk_buff *skb, __be16 protocol, 4809 int pulled_hlen, u16 mac_offset, 4810 int mac_len) 4811 { 4812 skb->protocol = protocol; 4813 skb->encapsulation = 1; 4814 skb_push(skb, pulled_hlen); 4815 skb_reset_transport_header(skb); 4816 skb->mac_header = mac_offset; 4817 skb->network_header = skb->mac_header + mac_len; 4818 skb->mac_len = mac_len; 4819 } 4820 4821 static inline bool netif_is_macsec(const struct net_device *dev) 4822 { 4823 return dev->priv_flags & IFF_MACSEC; 4824 } 4825 4826 static inline bool netif_is_macvlan(const struct net_device *dev) 4827 { 4828 return dev->priv_flags & IFF_MACVLAN; 4829 } 4830 4831 static inline bool netif_is_macvlan_port(const struct net_device *dev) 4832 { 4833 return dev->priv_flags & IFF_MACVLAN_PORT; 4834 } 4835 4836 static inline bool netif_is_bond_master(const struct net_device *dev) 4837 { 4838 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING; 4839 } 4840 4841 static inline bool netif_is_bond_slave(const struct net_device *dev) 4842 { 4843 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING; 4844 } 4845 4846 static inline bool netif_supports_nofcs(struct net_device *dev) 4847 { 4848 return dev->priv_flags & IFF_SUPP_NOFCS; 4849 } 4850 4851 static inline bool netif_has_l3_rx_handler(const struct net_device *dev) 4852 { 4853 return dev->priv_flags & IFF_L3MDEV_RX_HANDLER; 4854 } 4855 4856 static inline bool netif_is_l3_master(const struct net_device *dev) 4857 { 4858 return dev->priv_flags & IFF_L3MDEV_MASTER; 4859 } 4860 4861 static inline bool netif_is_l3_slave(const struct net_device *dev) 4862 { 4863 return dev->priv_flags & IFF_L3MDEV_SLAVE; 4864 } 4865 4866 static inline bool netif_is_bridge_master(const struct net_device *dev) 4867 { 4868 return dev->priv_flags & IFF_EBRIDGE; 4869 } 4870 4871 static inline bool netif_is_bridge_port(const struct net_device *dev) 4872 { 4873 return dev->priv_flags & IFF_BRIDGE_PORT; 4874 } 4875 4876 static inline bool netif_is_ovs_master(const struct net_device *dev) 4877 { 4878 return dev->priv_flags & IFF_OPENVSWITCH; 4879 } 4880 4881 static inline bool netif_is_ovs_port(const struct net_device *dev) 4882 { 4883 return dev->priv_flags & IFF_OVS_DATAPATH; 4884 } 4885 4886 static inline bool netif_is_any_bridge_port(const struct net_device *dev) 4887 { 4888 return netif_is_bridge_port(dev) || netif_is_ovs_port(dev); 4889 } 4890 4891 static inline bool netif_is_team_master(const struct net_device *dev) 4892 { 4893 return dev->priv_flags & IFF_TEAM; 4894 } 4895 4896 static inline bool netif_is_team_port(const struct net_device *dev) 4897 { 4898 return dev->priv_flags & IFF_TEAM_PORT; 4899 } 4900 4901 static inline bool netif_is_lag_master(const struct net_device *dev) 4902 { 4903 return netif_is_bond_master(dev) || netif_is_team_master(dev); 4904 } 4905 4906 static inline bool netif_is_lag_port(const struct net_device *dev) 4907 { 4908 return netif_is_bond_slave(dev) || netif_is_team_port(dev); 4909 } 4910 4911 static inline bool netif_is_rxfh_configured(const struct net_device *dev) 4912 { 4913 return dev->priv_flags & IFF_RXFH_CONFIGURED; 4914 } 4915 4916 static inline bool netif_is_failover(const struct net_device *dev) 4917 { 4918 return dev->priv_flags & IFF_FAILOVER; 4919 } 4920 4921 static inline bool netif_is_failover_slave(const struct net_device *dev) 4922 { 4923 return dev->priv_flags & IFF_FAILOVER_SLAVE; 4924 } 4925 4926 /* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */ 4927 static inline void netif_keep_dst(struct net_device *dev) 4928 { 4929 dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM); 4930 } 4931 4932 /* return true if dev can't cope with mtu frames that need vlan tag insertion */ 4933 static inline bool netif_reduces_vlan_mtu(struct net_device *dev) 4934 { 4935 /* TODO: reserve and use an additional IFF bit, if we get more users */ 4936 return netif_is_macsec(dev); 4937 } 4938 4939 extern struct pernet_operations __net_initdata loopback_net_ops; 4940 4941 /* Logging, debugging and troubleshooting/diagnostic helpers. */ 4942 4943 /* netdev_printk helpers, similar to dev_printk */ 4944 4945 static inline const char *netdev_name(const struct net_device *dev) 4946 { 4947 if (!dev->name[0] || strchr(dev->name, '%')) 4948 return "(unnamed net_device)"; 4949 return dev->name; 4950 } 4951 4952 static inline bool netdev_unregistering(const struct net_device *dev) 4953 { 4954 return dev->reg_state == NETREG_UNREGISTERING; 4955 } 4956 4957 static inline const char *netdev_reg_state(const struct net_device *dev) 4958 { 4959 switch (dev->reg_state) { 4960 case NETREG_UNINITIALIZED: return " (uninitialized)"; 4961 case NETREG_REGISTERED: return ""; 4962 case NETREG_UNREGISTERING: return " (unregistering)"; 4963 case NETREG_UNREGISTERED: return " (unregistered)"; 4964 case NETREG_RELEASED: return " (released)"; 4965 case NETREG_DUMMY: return " (dummy)"; 4966 } 4967 4968 WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, dev->reg_state); 4969 return " (unknown)"; 4970 } 4971 4972 __printf(3, 4) __cold 4973 void netdev_printk(const char *level, const struct net_device *dev, 4974 const char *format, ...); 4975 __printf(2, 3) __cold 4976 void netdev_emerg(const struct net_device *dev, const char *format, ...); 4977 __printf(2, 3) __cold 4978 void netdev_alert(const struct net_device *dev, const char *format, ...); 4979 __printf(2, 3) __cold 4980 void netdev_crit(const struct net_device *dev, const char *format, ...); 4981 __printf(2, 3) __cold 4982 void netdev_err(const struct net_device *dev, const char *format, ...); 4983 __printf(2, 3) __cold 4984 void netdev_warn(const struct net_device *dev, const char *format, ...); 4985 __printf(2, 3) __cold 4986 void netdev_notice(const struct net_device *dev, const char *format, ...); 4987 __printf(2, 3) __cold 4988 void netdev_info(const struct net_device *dev, const char *format, ...); 4989 4990 #define netdev_level_once(level, dev, fmt, ...) \ 4991 do { \ 4992 static bool __section(".data.once") __print_once; \ 4993 \ 4994 if (!__print_once) { \ 4995 __print_once = true; \ 4996 netdev_printk(level, dev, fmt, ##__VA_ARGS__); \ 4997 } \ 4998 } while (0) 4999 5000 #define netdev_emerg_once(dev, fmt, ...) \ 5001 netdev_level_once(KERN_EMERG, dev, fmt, ##__VA_ARGS__) 5002 #define netdev_alert_once(dev, fmt, ...) \ 5003 netdev_level_once(KERN_ALERT, dev, fmt, ##__VA_ARGS__) 5004 #define netdev_crit_once(dev, fmt, ...) \ 5005 netdev_level_once(KERN_CRIT, dev, fmt, ##__VA_ARGS__) 5006 #define netdev_err_once(dev, fmt, ...) \ 5007 netdev_level_once(KERN_ERR, dev, fmt, ##__VA_ARGS__) 5008 #define netdev_warn_once(dev, fmt, ...) \ 5009 netdev_level_once(KERN_WARNING, dev, fmt, ##__VA_ARGS__) 5010 #define netdev_notice_once(dev, fmt, ...) \ 5011 netdev_level_once(KERN_NOTICE, dev, fmt, ##__VA_ARGS__) 5012 #define netdev_info_once(dev, fmt, ...) \ 5013 netdev_level_once(KERN_INFO, dev, fmt, ##__VA_ARGS__) 5014 5015 #define MODULE_ALIAS_NETDEV(device) \ 5016 MODULE_ALIAS("netdev-" device) 5017 5018 #if defined(CONFIG_DYNAMIC_DEBUG) || \ 5019 (defined(CONFIG_DYNAMIC_DEBUG_CORE) && defined(DYNAMIC_DEBUG_MODULE)) 5020 #define netdev_dbg(__dev, format, args...) \ 5021 do { \ 5022 dynamic_netdev_dbg(__dev, format, ##args); \ 5023 } while (0) 5024 #elif defined(DEBUG) 5025 #define netdev_dbg(__dev, format, args...) \ 5026 netdev_printk(KERN_DEBUG, __dev, format, ##args) 5027 #else 5028 #define netdev_dbg(__dev, format, args...) \ 5029 ({ \ 5030 if (0) \ 5031 netdev_printk(KERN_DEBUG, __dev, format, ##args); \ 5032 }) 5033 #endif 5034 5035 #if defined(VERBOSE_DEBUG) 5036 #define netdev_vdbg netdev_dbg 5037 #else 5038 5039 #define netdev_vdbg(dev, format, args...) \ 5040 ({ \ 5041 if (0) \ 5042 netdev_printk(KERN_DEBUG, dev, format, ##args); \ 5043 0; \ 5044 }) 5045 #endif 5046 5047 /* 5048 * netdev_WARN() acts like dev_printk(), but with the key difference 5049 * of using a WARN/WARN_ON to get the message out, including the 5050 * file/line information and a backtrace. 5051 */ 5052 #define netdev_WARN(dev, format, args...) \ 5053 WARN(1, "netdevice: %s%s: " format, netdev_name(dev), \ 5054 netdev_reg_state(dev), ##args) 5055 5056 #define netdev_WARN_ONCE(dev, format, args...) \ 5057 WARN_ONCE(1, "netdevice: %s%s: " format, netdev_name(dev), \ 5058 netdev_reg_state(dev), ##args) 5059 5060 /* netif printk helpers, similar to netdev_printk */ 5061 5062 #define netif_printk(priv, type, level, dev, fmt, args...) \ 5063 do { \ 5064 if (netif_msg_##type(priv)) \ 5065 netdev_printk(level, (dev), fmt, ##args); \ 5066 } while (0) 5067 5068 #define netif_level(level, priv, type, dev, fmt, args...) \ 5069 do { \ 5070 if (netif_msg_##type(priv)) \ 5071 netdev_##level(dev, fmt, ##args); \ 5072 } while (0) 5073 5074 #define netif_emerg(priv, type, dev, fmt, args...) \ 5075 netif_level(emerg, priv, type, dev, fmt, ##args) 5076 #define netif_alert(priv, type, dev, fmt, args...) \ 5077 netif_level(alert, priv, type, dev, fmt, ##args) 5078 #define netif_crit(priv, type, dev, fmt, args...) \ 5079 netif_level(crit, priv, type, dev, fmt, ##args) 5080 #define netif_err(priv, type, dev, fmt, args...) \ 5081 netif_level(err, priv, type, dev, fmt, ##args) 5082 #define netif_warn(priv, type, dev, fmt, args...) \ 5083 netif_level(warn, priv, type, dev, fmt, ##args) 5084 #define netif_notice(priv, type, dev, fmt, args...) \ 5085 netif_level(notice, priv, type, dev, fmt, ##args) 5086 #define netif_info(priv, type, dev, fmt, args...) \ 5087 netif_level(info, priv, type, dev, fmt, ##args) 5088 5089 #if defined(CONFIG_DYNAMIC_DEBUG) || \ 5090 (defined(CONFIG_DYNAMIC_DEBUG_CORE) && defined(DYNAMIC_DEBUG_MODULE)) 5091 #define netif_dbg(priv, type, netdev, format, args...) \ 5092 do { \ 5093 if (netif_msg_##type(priv)) \ 5094 dynamic_netdev_dbg(netdev, format, ##args); \ 5095 } while (0) 5096 #elif defined(DEBUG) 5097 #define netif_dbg(priv, type, dev, format, args...) \ 5098 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args) 5099 #else 5100 #define netif_dbg(priv, type, dev, format, args...) \ 5101 ({ \ 5102 if (0) \ 5103 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \ 5104 0; \ 5105 }) 5106 #endif 5107 5108 /* if @cond then downgrade to debug, else print at @level */ 5109 #define netif_cond_dbg(priv, type, netdev, cond, level, fmt, args...) \ 5110 do { \ 5111 if (cond) \ 5112 netif_dbg(priv, type, netdev, fmt, ##args); \ 5113 else \ 5114 netif_ ## level(priv, type, netdev, fmt, ##args); \ 5115 } while (0) 5116 5117 #if defined(VERBOSE_DEBUG) 5118 #define netif_vdbg netif_dbg 5119 #else 5120 #define netif_vdbg(priv, type, dev, format, args...) \ 5121 ({ \ 5122 if (0) \ 5123 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \ 5124 0; \ 5125 }) 5126 #endif 5127 5128 /* 5129 * The list of packet types we will receive (as opposed to discard) 5130 * and the routines to invoke. 5131 * 5132 * Why 16. Because with 16 the only overlap we get on a hash of the 5133 * low nibble of the protocol value is RARP/SNAP/X.25. 5134 * 5135 * 0800 IP 5136 * 0001 802.3 5137 * 0002 AX.25 5138 * 0004 802.2 5139 * 8035 RARP 5140 * 0005 SNAP 5141 * 0805 X.25 5142 * 0806 ARP 5143 * 8137 IPX 5144 * 0009 Localtalk 5145 * 86DD IPv6 5146 */ 5147 #define PTYPE_HASH_SIZE (16) 5148 #define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1) 5149 5150 extern struct list_head ptype_all __read_mostly; 5151 extern struct list_head ptype_base[PTYPE_HASH_SIZE] __read_mostly; 5152 5153 extern struct net_device *blackhole_netdev; 5154 5155 #endif /* _LINUX_NETDEVICE_H */ 5156