1 /* SPDX-License-Identifier: GPL-2.0-or-later */ 2 /* 3 * INET An implementation of the TCP/IP protocol suite for the LINUX 4 * operating system. INET is implemented using the BSD Socket 5 * interface as the means of communication with the user level. 6 * 7 * Definitions for the Interfaces handler. 8 * 9 * Version: @(#)dev.h 1.0.10 08/12/93 10 * 11 * Authors: Ross Biro 12 * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG> 13 * Corey Minyard <wf-rch!minyard@relay.EU.net> 14 * Donald J. Becker, <becker@cesdis.gsfc.nasa.gov> 15 * Alan Cox, <alan@lxorguk.ukuu.org.uk> 16 * Bjorn Ekwall. <bj0rn@blox.se> 17 * Pekka Riikonen <priikone@poseidon.pspt.fi> 18 * 19 * Moved to /usr/include/linux for NET3 20 */ 21 #ifndef _LINUX_NETDEVICE_H 22 #define _LINUX_NETDEVICE_H 23 24 #include <linux/timer.h> 25 #include <linux/bug.h> 26 #include <linux/delay.h> 27 #include <linux/atomic.h> 28 #include <linux/prefetch.h> 29 #include <asm/cache.h> 30 #include <asm/byteorder.h> 31 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 <linux/ethtool.h> 38 #include <net/net_namespace.h> 39 #ifdef CONFIG_DCB 40 #include <net/dcbnl.h> 41 #endif 42 #include <net/netprio_cgroup.h> 43 #include <net/xdp.h> 44 45 #include <linux/netdev_features.h> 46 #include <linux/neighbour.h> 47 #include <uapi/linux/netdevice.h> 48 #include <uapi/linux/if_bonding.h> 49 #include <uapi/linux/pkt_cls.h> 50 #include <linux/hashtable.h> 51 52 struct netpoll_info; 53 struct device; 54 struct phy_device; 55 struct dsa_port; 56 57 struct sfp_bus; 58 /* 802.11 specific */ 59 struct wireless_dev; 60 /* 802.15.4 specific */ 61 struct wpan_dev; 62 struct mpls_dev; 63 /* UDP Tunnel offloads */ 64 struct udp_tunnel_info; 65 struct bpf_prog; 66 struct xdp_buff; 67 68 void netdev_set_default_ethtool_ops(struct net_device *dev, 69 const struct ethtool_ops *ops); 70 71 /* Backlog congestion levels */ 72 #define NET_RX_SUCCESS 0 /* keep 'em coming, baby */ 73 #define NET_RX_DROP 1 /* packet dropped */ 74 75 /* 76 * Transmit return codes: transmit return codes originate from three different 77 * namespaces: 78 * 79 * - qdisc return codes 80 * - driver transmit return codes 81 * - errno values 82 * 83 * Drivers are allowed to return any one of those in their hard_start_xmit() 84 * function. Real network devices commonly used with qdiscs should only return 85 * the driver transmit return codes though - when qdiscs are used, the actual 86 * transmission happens asynchronously, so the value is not propagated to 87 * higher layers. Virtual network devices transmit synchronously; in this case 88 * the driver transmit return codes are consumed by dev_queue_xmit(), and all 89 * others are propagated to higher layers. 90 */ 91 92 /* qdisc ->enqueue() return codes. */ 93 #define NET_XMIT_SUCCESS 0x00 94 #define NET_XMIT_DROP 0x01 /* skb dropped */ 95 #define NET_XMIT_CN 0x02 /* congestion notification */ 96 #define NET_XMIT_MASK 0x0f /* qdisc flags in net/sch_generic.h */ 97 98 /* NET_XMIT_CN is special. It does not guarantee that this packet is lost. It 99 * indicates that the device will soon be dropping packets, or already drops 100 * some packets of the same priority; prompting us to send less aggressively. */ 101 #define net_xmit_eval(e) ((e) == NET_XMIT_CN ? 0 : (e)) 102 #define net_xmit_errno(e) ((e) != NET_XMIT_CN ? -ENOBUFS : 0) 103 104 /* Driver transmit return codes */ 105 #define NETDEV_TX_MASK 0xf0 106 107 enum netdev_tx { 108 __NETDEV_TX_MIN = INT_MIN, /* make sure enum is signed */ 109 NETDEV_TX_OK = 0x00, /* driver took care of packet */ 110 NETDEV_TX_BUSY = 0x10, /* driver tx path was busy*/ 111 }; 112 typedef enum netdev_tx netdev_tx_t; 113 114 /* 115 * Current order: NETDEV_TX_MASK > NET_XMIT_MASK >= 0 is significant; 116 * hard_start_xmit() return < NET_XMIT_MASK means skb was consumed. 117 */ 118 static inline bool dev_xmit_complete(int rc) 119 { 120 /* 121 * Positive cases with an skb consumed by a driver: 122 * - successful transmission (rc == NETDEV_TX_OK) 123 * - error while transmitting (rc < 0) 124 * - error while queueing to a different device (rc & NET_XMIT_MASK) 125 */ 126 if (likely(rc < NET_XMIT_MASK)) 127 return true; 128 129 return false; 130 } 131 132 /* 133 * Compute the worst-case header length according to the protocols 134 * used. 135 */ 136 137 #if defined(CONFIG_HYPERV_NET) 138 # define LL_MAX_HEADER 128 139 #elif defined(CONFIG_WLAN) || IS_ENABLED(CONFIG_AX25) 140 # if defined(CONFIG_MAC80211_MESH) 141 # define LL_MAX_HEADER 128 142 # else 143 # define LL_MAX_HEADER 96 144 # endif 145 #else 146 # define LL_MAX_HEADER 32 147 #endif 148 149 #if !IS_ENABLED(CONFIG_NET_IPIP) && !IS_ENABLED(CONFIG_NET_IPGRE) && \ 150 !IS_ENABLED(CONFIG_IPV6_SIT) && !IS_ENABLED(CONFIG_IPV6_TUNNEL) 151 #define MAX_HEADER LL_MAX_HEADER 152 #else 153 #define MAX_HEADER (LL_MAX_HEADER + 48) 154 #endif 155 156 /* 157 * Old network device statistics. Fields are native words 158 * (unsigned long) so they can be read and written atomically. 159 */ 160 161 struct net_device_stats { 162 unsigned long rx_packets; 163 unsigned long tx_packets; 164 unsigned long rx_bytes; 165 unsigned long tx_bytes; 166 unsigned long rx_errors; 167 unsigned long tx_errors; 168 unsigned long rx_dropped; 169 unsigned long tx_dropped; 170 unsigned long multicast; 171 unsigned long collisions; 172 unsigned long rx_length_errors; 173 unsigned long rx_over_errors; 174 unsigned long rx_crc_errors; 175 unsigned long rx_frame_errors; 176 unsigned long rx_fifo_errors; 177 unsigned long rx_missed_errors; 178 unsigned long tx_aborted_errors; 179 unsigned long tx_carrier_errors; 180 unsigned long tx_fifo_errors; 181 unsigned long tx_heartbeat_errors; 182 unsigned long tx_window_errors; 183 unsigned long rx_compressed; 184 unsigned long tx_compressed; 185 }; 186 187 188 #include <linux/cache.h> 189 #include <linux/skbuff.h> 190 191 #ifdef CONFIG_RPS 192 #include <linux/static_key.h> 193 extern struct static_key_false rps_needed; 194 extern struct static_key_false rfs_needed; 195 #endif 196 197 struct neighbour; 198 struct neigh_parms; 199 struct sk_buff; 200 201 struct netdev_hw_addr { 202 struct list_head list; 203 unsigned char addr[MAX_ADDR_LEN]; 204 unsigned char type; 205 #define NETDEV_HW_ADDR_T_LAN 1 206 #define NETDEV_HW_ADDR_T_SAN 2 207 #define NETDEV_HW_ADDR_T_SLAVE 3 208 #define NETDEV_HW_ADDR_T_UNICAST 4 209 #define NETDEV_HW_ADDR_T_MULTICAST 5 210 bool global_use; 211 int sync_cnt; 212 int refcount; 213 int synced; 214 struct rcu_head rcu_head; 215 }; 216 217 struct netdev_hw_addr_list { 218 struct list_head list; 219 int count; 220 }; 221 222 #define netdev_hw_addr_list_count(l) ((l)->count) 223 #define netdev_hw_addr_list_empty(l) (netdev_hw_addr_list_count(l) == 0) 224 #define netdev_hw_addr_list_for_each(ha, l) \ 225 list_for_each_entry(ha, &(l)->list, list) 226 227 #define netdev_uc_count(dev) netdev_hw_addr_list_count(&(dev)->uc) 228 #define netdev_uc_empty(dev) netdev_hw_addr_list_empty(&(dev)->uc) 229 #define netdev_for_each_uc_addr(ha, dev) \ 230 netdev_hw_addr_list_for_each(ha, &(dev)->uc) 231 232 #define netdev_mc_count(dev) netdev_hw_addr_list_count(&(dev)->mc) 233 #define netdev_mc_empty(dev) netdev_hw_addr_list_empty(&(dev)->mc) 234 #define netdev_for_each_mc_addr(ha, dev) \ 235 netdev_hw_addr_list_for_each(ha, &(dev)->mc) 236 237 struct hh_cache { 238 unsigned int hh_len; 239 seqlock_t hh_lock; 240 241 /* cached hardware header; allow for machine alignment needs. */ 242 #define HH_DATA_MOD 16 243 #define HH_DATA_OFF(__len) \ 244 (HH_DATA_MOD - (((__len - 1) & (HH_DATA_MOD - 1)) + 1)) 245 #define HH_DATA_ALIGN(__len) \ 246 (((__len)+(HH_DATA_MOD-1))&~(HH_DATA_MOD - 1)) 247 unsigned long hh_data[HH_DATA_ALIGN(LL_MAX_HEADER) / sizeof(long)]; 248 }; 249 250 /* Reserve HH_DATA_MOD byte-aligned hard_header_len, but at least that much. 251 * Alternative is: 252 * dev->hard_header_len ? (dev->hard_header_len + 253 * (HH_DATA_MOD - 1)) & ~(HH_DATA_MOD - 1) : 0 254 * 255 * We could use other alignment values, but we must maintain the 256 * relationship HH alignment <= LL alignment. 257 */ 258 #define LL_RESERVED_SPACE(dev) \ 259 ((((dev)->hard_header_len+(dev)->needed_headroom)&~(HH_DATA_MOD - 1)) + HH_DATA_MOD) 260 #define LL_RESERVED_SPACE_EXTRA(dev,extra) \ 261 ((((dev)->hard_header_len+(dev)->needed_headroom+(extra))&~(HH_DATA_MOD - 1)) + HH_DATA_MOD) 262 263 struct header_ops { 264 int (*create) (struct sk_buff *skb, struct net_device *dev, 265 unsigned short type, const void *daddr, 266 const void *saddr, unsigned int len); 267 int (*parse)(const struct sk_buff *skb, unsigned char *haddr); 268 int (*cache)(const struct neighbour *neigh, struct hh_cache *hh, __be16 type); 269 void (*cache_update)(struct hh_cache *hh, 270 const struct net_device *dev, 271 const unsigned char *haddr); 272 bool (*validate)(const char *ll_header, unsigned int len); 273 __be16 (*parse_protocol)(const struct sk_buff *skb); 274 }; 275 276 /* These flag bits are private to the generic network queueing 277 * layer; they may not be explicitly referenced by any other 278 * code. 279 */ 280 281 enum netdev_state_t { 282 __LINK_STATE_START, 283 __LINK_STATE_PRESENT, 284 __LINK_STATE_NOCARRIER, 285 __LINK_STATE_LINKWATCH_PENDING, 286 __LINK_STATE_DORMANT, 287 }; 288 289 290 /* 291 * This structure holds boot-time configured netdevice settings. They 292 * are then used in the device probing. 293 */ 294 struct netdev_boot_setup { 295 char name[IFNAMSIZ]; 296 struct ifmap map; 297 }; 298 #define NETDEV_BOOT_SETUP_MAX 8 299 300 int __init netdev_boot_setup(char *str); 301 302 struct gro_list { 303 struct list_head list; 304 int count; 305 }; 306 307 /* 308 * size of gro hash buckets, must less than bit number of 309 * napi_struct::gro_bitmask 310 */ 311 #define GRO_HASH_BUCKETS 8 312 313 /* 314 * Structure for NAPI scheduling similar to tasklet but with weighting 315 */ 316 struct napi_struct { 317 /* The poll_list must only be managed by the entity which 318 * changes the state of the NAPI_STATE_SCHED bit. This means 319 * whoever atomically sets that bit can add this napi_struct 320 * to the per-CPU poll_list, and whoever clears that bit 321 * can remove from the list right before clearing the bit. 322 */ 323 struct list_head poll_list; 324 325 unsigned long state; 326 int weight; 327 unsigned long gro_bitmask; 328 int (*poll)(struct napi_struct *, int); 329 #ifdef CONFIG_NETPOLL 330 int poll_owner; 331 #endif 332 struct net_device *dev; 333 struct gro_list gro_hash[GRO_HASH_BUCKETS]; 334 struct sk_buff *skb; 335 struct list_head rx_list; /* Pending GRO_NORMAL skbs */ 336 int rx_count; /* length of rx_list */ 337 struct hrtimer timer; 338 struct list_head dev_list; 339 struct hlist_node napi_hash_node; 340 unsigned int napi_id; 341 }; 342 343 enum { 344 NAPI_STATE_SCHED, /* Poll is scheduled */ 345 NAPI_STATE_MISSED, /* reschedule a napi */ 346 NAPI_STATE_DISABLE, /* Disable pending */ 347 NAPI_STATE_NPSVC, /* Netpoll - don't dequeue from poll_list */ 348 NAPI_STATE_HASHED, /* In NAPI hash (busy polling possible) */ 349 NAPI_STATE_NO_BUSY_POLL,/* Do not add in napi_hash, no busy polling */ 350 NAPI_STATE_IN_BUSY_POLL,/* sk_busy_loop() owns this NAPI */ 351 }; 352 353 enum { 354 NAPIF_STATE_SCHED = BIT(NAPI_STATE_SCHED), 355 NAPIF_STATE_MISSED = BIT(NAPI_STATE_MISSED), 356 NAPIF_STATE_DISABLE = BIT(NAPI_STATE_DISABLE), 357 NAPIF_STATE_NPSVC = BIT(NAPI_STATE_NPSVC), 358 NAPIF_STATE_HASHED = BIT(NAPI_STATE_HASHED), 359 NAPIF_STATE_NO_BUSY_POLL = BIT(NAPI_STATE_NO_BUSY_POLL), 360 NAPIF_STATE_IN_BUSY_POLL = BIT(NAPI_STATE_IN_BUSY_POLL), 361 }; 362 363 enum gro_result { 364 GRO_MERGED, 365 GRO_MERGED_FREE, 366 GRO_HELD, 367 GRO_NORMAL, 368 GRO_DROP, 369 GRO_CONSUMED, 370 }; 371 typedef enum gro_result gro_result_t; 372 373 /* 374 * enum rx_handler_result - Possible return values for rx_handlers. 375 * @RX_HANDLER_CONSUMED: skb was consumed by rx_handler, do not process it 376 * further. 377 * @RX_HANDLER_ANOTHER: Do another round in receive path. This is indicated in 378 * case skb->dev was changed by rx_handler. 379 * @RX_HANDLER_EXACT: Force exact delivery, no wildcard. 380 * @RX_HANDLER_PASS: Do nothing, pass the skb as if no rx_handler was called. 381 * 382 * rx_handlers are functions called from inside __netif_receive_skb(), to do 383 * special processing of the skb, prior to delivery to protocol handlers. 384 * 385 * Currently, a net_device can only have a single rx_handler registered. Trying 386 * to register a second rx_handler will return -EBUSY. 387 * 388 * To register a rx_handler on a net_device, use netdev_rx_handler_register(). 389 * To unregister a rx_handler on a net_device, use 390 * netdev_rx_handler_unregister(). 391 * 392 * Upon return, rx_handler is expected to tell __netif_receive_skb() what to 393 * do with the skb. 394 * 395 * If the rx_handler consumed the skb in some way, it should return 396 * RX_HANDLER_CONSUMED. This is appropriate when the rx_handler arranged for 397 * the skb to be delivered in some other way. 398 * 399 * If the rx_handler changed skb->dev, to divert the skb to another 400 * net_device, it should return RX_HANDLER_ANOTHER. The rx_handler for the 401 * new device will be called if it exists. 402 * 403 * If the rx_handler decides the skb should be ignored, it should return 404 * RX_HANDLER_EXACT. The skb will only be delivered to protocol handlers that 405 * are registered on exact device (ptype->dev == skb->dev). 406 * 407 * If the rx_handler didn't change skb->dev, but wants the skb to be normally 408 * delivered, it should return RX_HANDLER_PASS. 409 * 410 * A device without a registered rx_handler will behave as if rx_handler 411 * returned RX_HANDLER_PASS. 412 */ 413 414 enum rx_handler_result { 415 RX_HANDLER_CONSUMED, 416 RX_HANDLER_ANOTHER, 417 RX_HANDLER_EXACT, 418 RX_HANDLER_PASS, 419 }; 420 typedef enum rx_handler_result rx_handler_result_t; 421 typedef rx_handler_result_t rx_handler_func_t(struct sk_buff **pskb); 422 423 void __napi_schedule(struct napi_struct *n); 424 void __napi_schedule_irqoff(struct napi_struct *n); 425 426 static inline bool napi_disable_pending(struct napi_struct *n) 427 { 428 return test_bit(NAPI_STATE_DISABLE, &n->state); 429 } 430 431 bool napi_schedule_prep(struct napi_struct *n); 432 433 /** 434 * napi_schedule - schedule NAPI poll 435 * @n: NAPI context 436 * 437 * Schedule NAPI poll routine to be called if it is not already 438 * running. 439 */ 440 static inline void napi_schedule(struct napi_struct *n) 441 { 442 if (napi_schedule_prep(n)) 443 __napi_schedule(n); 444 } 445 446 /** 447 * napi_schedule_irqoff - schedule NAPI poll 448 * @n: NAPI context 449 * 450 * Variant of napi_schedule(), assuming hard irqs are masked. 451 */ 452 static inline void napi_schedule_irqoff(struct napi_struct *n) 453 { 454 if (napi_schedule_prep(n)) 455 __napi_schedule_irqoff(n); 456 } 457 458 /* Try to reschedule poll. Called by dev->poll() after napi_complete(). */ 459 static inline bool napi_reschedule(struct napi_struct *napi) 460 { 461 if (napi_schedule_prep(napi)) { 462 __napi_schedule(napi); 463 return true; 464 } 465 return false; 466 } 467 468 bool napi_complete_done(struct napi_struct *n, int work_done); 469 /** 470 * napi_complete - NAPI processing complete 471 * @n: NAPI context 472 * 473 * Mark NAPI processing as complete. 474 * Consider using napi_complete_done() instead. 475 * Return false if device should avoid rearming interrupts. 476 */ 477 static inline bool napi_complete(struct napi_struct *n) 478 { 479 return napi_complete_done(n, 0); 480 } 481 482 /** 483 * napi_hash_del - remove a NAPI from global table 484 * @napi: NAPI context 485 * 486 * Warning: caller must observe RCU grace period 487 * before freeing memory containing @napi, if 488 * this function returns true. 489 * Note: core networking stack automatically calls it 490 * from netif_napi_del(). 491 * Drivers might want to call this helper to combine all 492 * the needed RCU grace periods into a single one. 493 */ 494 bool napi_hash_del(struct napi_struct *napi); 495 496 /** 497 * napi_disable - prevent NAPI from scheduling 498 * @n: NAPI context 499 * 500 * Stop NAPI from being scheduled on this context. 501 * Waits till any outstanding processing completes. 502 */ 503 void napi_disable(struct napi_struct *n); 504 505 /** 506 * napi_enable - enable NAPI scheduling 507 * @n: NAPI context 508 * 509 * Resume NAPI from being scheduled on this context. 510 * Must be paired with napi_disable. 511 */ 512 static inline void napi_enable(struct napi_struct *n) 513 { 514 BUG_ON(!test_bit(NAPI_STATE_SCHED, &n->state)); 515 smp_mb__before_atomic(); 516 clear_bit(NAPI_STATE_SCHED, &n->state); 517 clear_bit(NAPI_STATE_NPSVC, &n->state); 518 } 519 520 /** 521 * napi_synchronize - wait until NAPI is not running 522 * @n: NAPI context 523 * 524 * Wait until NAPI is done being scheduled on this context. 525 * Waits till any outstanding processing completes but 526 * does not disable future activations. 527 */ 528 static inline void napi_synchronize(const struct napi_struct *n) 529 { 530 if (IS_ENABLED(CONFIG_SMP)) 531 while (test_bit(NAPI_STATE_SCHED, &n->state)) 532 msleep(1); 533 else 534 barrier(); 535 } 536 537 /** 538 * napi_if_scheduled_mark_missed - if napi is running, set the 539 * NAPIF_STATE_MISSED 540 * @n: NAPI context 541 * 542 * If napi is running, set the NAPIF_STATE_MISSED, and return true if 543 * NAPI is scheduled. 544 **/ 545 static inline bool napi_if_scheduled_mark_missed(struct napi_struct *n) 546 { 547 unsigned long val, new; 548 549 do { 550 val = READ_ONCE(n->state); 551 if (val & NAPIF_STATE_DISABLE) 552 return true; 553 554 if (!(val & NAPIF_STATE_SCHED)) 555 return false; 556 557 new = val | NAPIF_STATE_MISSED; 558 } while (cmpxchg(&n->state, val, new) != val); 559 560 return true; 561 } 562 563 enum netdev_queue_state_t { 564 __QUEUE_STATE_DRV_XOFF, 565 __QUEUE_STATE_STACK_XOFF, 566 __QUEUE_STATE_FROZEN, 567 }; 568 569 #define QUEUE_STATE_DRV_XOFF (1 << __QUEUE_STATE_DRV_XOFF) 570 #define QUEUE_STATE_STACK_XOFF (1 << __QUEUE_STATE_STACK_XOFF) 571 #define QUEUE_STATE_FROZEN (1 << __QUEUE_STATE_FROZEN) 572 573 #define QUEUE_STATE_ANY_XOFF (QUEUE_STATE_DRV_XOFF | QUEUE_STATE_STACK_XOFF) 574 #define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \ 575 QUEUE_STATE_FROZEN) 576 #define QUEUE_STATE_DRV_XOFF_OR_FROZEN (QUEUE_STATE_DRV_XOFF | \ 577 QUEUE_STATE_FROZEN) 578 579 /* 580 * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue. The 581 * netif_tx_* functions below are used to manipulate this flag. The 582 * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit 583 * queue independently. The netif_xmit_*stopped functions below are called 584 * to check if the queue has been stopped by the driver or stack (either 585 * of the XOFF bits are set in the state). Drivers should not need to call 586 * netif_xmit*stopped functions, they should only be using netif_tx_*. 587 */ 588 589 struct netdev_queue { 590 /* 591 * read-mostly part 592 */ 593 struct net_device *dev; 594 struct Qdisc __rcu *qdisc; 595 struct Qdisc *qdisc_sleeping; 596 #ifdef CONFIG_SYSFS 597 struct kobject kobj; 598 #endif 599 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA) 600 int numa_node; 601 #endif 602 unsigned long tx_maxrate; 603 /* 604 * Number of TX timeouts for this queue 605 * (/sys/class/net/DEV/Q/trans_timeout) 606 */ 607 unsigned long trans_timeout; 608 609 /* Subordinate device that the queue has been assigned to */ 610 struct net_device *sb_dev; 611 #ifdef CONFIG_XDP_SOCKETS 612 struct xdp_umem *umem; 613 #endif 614 /* 615 * write-mostly part 616 */ 617 spinlock_t _xmit_lock ____cacheline_aligned_in_smp; 618 int xmit_lock_owner; 619 /* 620 * Time (in jiffies) of last Tx 621 */ 622 unsigned long trans_start; 623 624 unsigned long state; 625 626 #ifdef CONFIG_BQL 627 struct dql dql; 628 #endif 629 } ____cacheline_aligned_in_smp; 630 631 extern int sysctl_fb_tunnels_only_for_init_net; 632 extern int sysctl_devconf_inherit_init_net; 633 634 static inline bool net_has_fallback_tunnels(const struct net *net) 635 { 636 return net == &init_net || 637 !IS_ENABLED(CONFIG_SYSCTL) || 638 !sysctl_fb_tunnels_only_for_init_net; 639 } 640 641 static inline int netdev_queue_numa_node_read(const struct netdev_queue *q) 642 { 643 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA) 644 return q->numa_node; 645 #else 646 return NUMA_NO_NODE; 647 #endif 648 } 649 650 static inline void netdev_queue_numa_node_write(struct netdev_queue *q, int node) 651 { 652 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA) 653 q->numa_node = node; 654 #endif 655 } 656 657 #ifdef CONFIG_RPS 658 /* 659 * This structure holds an RPS map which can be of variable length. The 660 * map is an array of CPUs. 661 */ 662 struct rps_map { 663 unsigned int len; 664 struct rcu_head rcu; 665 u16 cpus[0]; 666 }; 667 #define RPS_MAP_SIZE(_num) (sizeof(struct rps_map) + ((_num) * sizeof(u16))) 668 669 /* 670 * The rps_dev_flow structure contains the mapping of a flow to a CPU, the 671 * tail pointer for that CPU's input queue at the time of last enqueue, and 672 * a hardware filter index. 673 */ 674 struct rps_dev_flow { 675 u16 cpu; 676 u16 filter; 677 unsigned int last_qtail; 678 }; 679 #define RPS_NO_FILTER 0xffff 680 681 /* 682 * The rps_dev_flow_table structure contains a table of flow mappings. 683 */ 684 struct rps_dev_flow_table { 685 unsigned int mask; 686 struct rcu_head rcu; 687 struct rps_dev_flow flows[0]; 688 }; 689 #define RPS_DEV_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_dev_flow_table) + \ 690 ((_num) * sizeof(struct rps_dev_flow))) 691 692 /* 693 * The rps_sock_flow_table contains mappings of flows to the last CPU 694 * on which they were processed by the application (set in recvmsg). 695 * Each entry is a 32bit value. Upper part is the high-order bits 696 * of flow hash, lower part is CPU number. 697 * rps_cpu_mask is used to partition the space, depending on number of 698 * possible CPUs : rps_cpu_mask = roundup_pow_of_two(nr_cpu_ids) - 1 699 * For example, if 64 CPUs are possible, rps_cpu_mask = 0x3f, 700 * meaning we use 32-6=26 bits for the hash. 701 */ 702 struct rps_sock_flow_table { 703 u32 mask; 704 705 u32 ents[0] ____cacheline_aligned_in_smp; 706 }; 707 #define RPS_SOCK_FLOW_TABLE_SIZE(_num) (offsetof(struct rps_sock_flow_table, ents[_num])) 708 709 #define RPS_NO_CPU 0xffff 710 711 extern u32 rps_cpu_mask; 712 extern struct rps_sock_flow_table __rcu *rps_sock_flow_table; 713 714 static inline void rps_record_sock_flow(struct rps_sock_flow_table *table, 715 u32 hash) 716 { 717 if (table && hash) { 718 unsigned int index = hash & table->mask; 719 u32 val = hash & ~rps_cpu_mask; 720 721 /* We only give a hint, preemption can change CPU under us */ 722 val |= raw_smp_processor_id(); 723 724 if (table->ents[index] != val) 725 table->ents[index] = val; 726 } 727 } 728 729 #ifdef CONFIG_RFS_ACCEL 730 bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index, u32 flow_id, 731 u16 filter_id); 732 #endif 733 #endif /* CONFIG_RPS */ 734 735 /* This structure contains an instance of an RX queue. */ 736 struct netdev_rx_queue { 737 #ifdef CONFIG_RPS 738 struct rps_map __rcu *rps_map; 739 struct rps_dev_flow_table __rcu *rps_flow_table; 740 #endif 741 struct kobject kobj; 742 struct net_device *dev; 743 struct xdp_rxq_info xdp_rxq; 744 #ifdef CONFIG_XDP_SOCKETS 745 struct xdp_umem *umem; 746 #endif 747 } ____cacheline_aligned_in_smp; 748 749 /* 750 * RX queue sysfs structures and functions. 751 */ 752 struct rx_queue_attribute { 753 struct attribute attr; 754 ssize_t (*show)(struct netdev_rx_queue *queue, char *buf); 755 ssize_t (*store)(struct netdev_rx_queue *queue, 756 const char *buf, size_t len); 757 }; 758 759 #ifdef CONFIG_XPS 760 /* 761 * This structure holds an XPS map which can be of variable length. The 762 * map is an array of queues. 763 */ 764 struct xps_map { 765 unsigned int len; 766 unsigned int alloc_len; 767 struct rcu_head rcu; 768 u16 queues[0]; 769 }; 770 #define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16))) 771 #define XPS_MIN_MAP_ALLOC ((L1_CACHE_ALIGN(offsetof(struct xps_map, queues[1])) \ 772 - sizeof(struct xps_map)) / sizeof(u16)) 773 774 /* 775 * This structure holds all XPS maps for device. Maps are indexed by CPU. 776 */ 777 struct xps_dev_maps { 778 struct rcu_head rcu; 779 struct xps_map __rcu *attr_map[0]; /* Either CPUs map or RXQs map */ 780 }; 781 782 #define XPS_CPU_DEV_MAPS_SIZE(_tcs) (sizeof(struct xps_dev_maps) + \ 783 (nr_cpu_ids * (_tcs) * sizeof(struct xps_map *))) 784 785 #define XPS_RXQ_DEV_MAPS_SIZE(_tcs, _rxqs) (sizeof(struct xps_dev_maps) +\ 786 (_rxqs * (_tcs) * sizeof(struct xps_map *))) 787 788 #endif /* CONFIG_XPS */ 789 790 #define TC_MAX_QUEUE 16 791 #define TC_BITMASK 15 792 /* HW offloaded queuing disciplines txq count and offset maps */ 793 struct netdev_tc_txq { 794 u16 count; 795 u16 offset; 796 }; 797 798 #if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE) 799 /* 800 * This structure is to hold information about the device 801 * configured to run FCoE protocol stack. 802 */ 803 struct netdev_fcoe_hbainfo { 804 char manufacturer[64]; 805 char serial_number[64]; 806 char hardware_version[64]; 807 char driver_version[64]; 808 char optionrom_version[64]; 809 char firmware_version[64]; 810 char model[256]; 811 char model_description[256]; 812 }; 813 #endif 814 815 #define MAX_PHYS_ITEM_ID_LEN 32 816 817 /* This structure holds a unique identifier to identify some 818 * physical item (port for example) used by a netdevice. 819 */ 820 struct netdev_phys_item_id { 821 unsigned char id[MAX_PHYS_ITEM_ID_LEN]; 822 unsigned char id_len; 823 }; 824 825 static inline bool netdev_phys_item_id_same(struct netdev_phys_item_id *a, 826 struct netdev_phys_item_id *b) 827 { 828 return a->id_len == b->id_len && 829 memcmp(a->id, b->id, a->id_len) == 0; 830 } 831 832 typedef u16 (*select_queue_fallback_t)(struct net_device *dev, 833 struct sk_buff *skb, 834 struct net_device *sb_dev); 835 836 enum tc_setup_type { 837 TC_SETUP_QDISC_MQPRIO, 838 TC_SETUP_CLSU32, 839 TC_SETUP_CLSFLOWER, 840 TC_SETUP_CLSMATCHALL, 841 TC_SETUP_CLSBPF, 842 TC_SETUP_BLOCK, 843 TC_SETUP_QDISC_CBS, 844 TC_SETUP_QDISC_RED, 845 TC_SETUP_QDISC_PRIO, 846 TC_SETUP_QDISC_MQ, 847 TC_SETUP_QDISC_ETF, 848 TC_SETUP_ROOT_QDISC, 849 TC_SETUP_QDISC_GRED, 850 TC_SETUP_QDISC_TAPRIO, 851 TC_SETUP_FT, 852 }; 853 854 /* These structures hold the attributes of bpf state that are being passed 855 * to the netdevice through the bpf op. 856 */ 857 enum bpf_netdev_command { 858 /* Set or clear a bpf program used in the earliest stages of packet 859 * rx. The prog will have been loaded as BPF_PROG_TYPE_XDP. The callee 860 * is responsible for calling bpf_prog_put on any old progs that are 861 * stored. In case of error, the callee need not release the new prog 862 * reference, but on success it takes ownership and must bpf_prog_put 863 * when it is no longer used. 864 */ 865 XDP_SETUP_PROG, 866 XDP_SETUP_PROG_HW, 867 XDP_QUERY_PROG, 868 XDP_QUERY_PROG_HW, 869 /* BPF program for offload callbacks, invoked at program load time. */ 870 BPF_OFFLOAD_MAP_ALLOC, 871 BPF_OFFLOAD_MAP_FREE, 872 XDP_SETUP_XSK_UMEM, 873 }; 874 875 struct bpf_prog_offload_ops; 876 struct netlink_ext_ack; 877 struct xdp_umem; 878 879 struct netdev_bpf { 880 enum bpf_netdev_command command; 881 union { 882 /* XDP_SETUP_PROG */ 883 struct { 884 u32 flags; 885 struct bpf_prog *prog; 886 struct netlink_ext_ack *extack; 887 }; 888 /* XDP_QUERY_PROG, XDP_QUERY_PROG_HW */ 889 struct { 890 u32 prog_id; 891 /* flags with which program was installed */ 892 u32 prog_flags; 893 }; 894 /* BPF_OFFLOAD_MAP_ALLOC, BPF_OFFLOAD_MAP_FREE */ 895 struct { 896 struct bpf_offloaded_map *offmap; 897 }; 898 /* XDP_SETUP_XSK_UMEM */ 899 struct { 900 struct xdp_umem *umem; 901 u16 queue_id; 902 } xsk; 903 }; 904 }; 905 906 /* Flags for ndo_xsk_wakeup. */ 907 #define XDP_WAKEUP_RX (1 << 0) 908 #define XDP_WAKEUP_TX (1 << 1) 909 910 #ifdef CONFIG_XFRM_OFFLOAD 911 struct xfrmdev_ops { 912 int (*xdo_dev_state_add) (struct xfrm_state *x); 913 void (*xdo_dev_state_delete) (struct xfrm_state *x); 914 void (*xdo_dev_state_free) (struct xfrm_state *x); 915 bool (*xdo_dev_offload_ok) (struct sk_buff *skb, 916 struct xfrm_state *x); 917 void (*xdo_dev_state_advance_esn) (struct xfrm_state *x); 918 }; 919 #endif 920 921 struct dev_ifalias { 922 struct rcu_head rcuhead; 923 char ifalias[]; 924 }; 925 926 struct devlink; 927 struct tlsdev_ops; 928 929 struct netdev_name_node { 930 struct hlist_node hlist; 931 struct list_head list; 932 struct net_device *dev; 933 const char *name; 934 }; 935 936 int netdev_name_node_alt_create(struct net_device *dev, const char *name); 937 int netdev_name_node_alt_destroy(struct net_device *dev, const char *name); 938 939 /* 940 * This structure defines the management hooks for network devices. 941 * The following hooks can be defined; unless noted otherwise, they are 942 * optional and can be filled with a null pointer. 943 * 944 * int (*ndo_init)(struct net_device *dev); 945 * This function is called once when a network device is registered. 946 * The network device can use this for any late stage initialization 947 * or semantic validation. It can fail with an error code which will 948 * be propagated back to register_netdev. 949 * 950 * void (*ndo_uninit)(struct net_device *dev); 951 * This function is called when device is unregistered or when registration 952 * fails. It is not called if init fails. 953 * 954 * int (*ndo_open)(struct net_device *dev); 955 * This function is called when a network device transitions to the up 956 * state. 957 * 958 * int (*ndo_stop)(struct net_device *dev); 959 * This function is called when a network device transitions to the down 960 * state. 961 * 962 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb, 963 * struct net_device *dev); 964 * Called when a packet needs to be transmitted. 965 * Returns NETDEV_TX_OK. Can return NETDEV_TX_BUSY, but you should stop 966 * the queue before that can happen; it's for obsolete devices and weird 967 * corner cases, but the stack really does a non-trivial amount 968 * of useless work if you return NETDEV_TX_BUSY. 969 * Required; cannot be NULL. 970 * 971 * netdev_features_t (*ndo_features_check)(struct sk_buff *skb, 972 * struct net_device *dev 973 * netdev_features_t features); 974 * Called by core transmit path to determine if device is capable of 975 * performing offload operations on a given packet. This is to give 976 * the device an opportunity to implement any restrictions that cannot 977 * be otherwise expressed by feature flags. The check is called with 978 * the set of features that the stack has calculated and it returns 979 * those the driver believes to be appropriate. 980 * 981 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb, 982 * struct net_device *sb_dev); 983 * Called to decide which queue to use when device supports multiple 984 * transmit queues. 985 * 986 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags); 987 * This function is called to allow device receiver to make 988 * changes to configuration when multicast or promiscuous is enabled. 989 * 990 * void (*ndo_set_rx_mode)(struct net_device *dev); 991 * This function is called device changes address list filtering. 992 * If driver handles unicast address filtering, it should set 993 * IFF_UNICAST_FLT in its priv_flags. 994 * 995 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr); 996 * This function is called when the Media Access Control address 997 * needs to be changed. If this interface is not defined, the 998 * MAC address can not be changed. 999 * 1000 * int (*ndo_validate_addr)(struct net_device *dev); 1001 * Test if Media Access Control address is valid for the device. 1002 * 1003 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd); 1004 * Called when a user requests an ioctl which can't be handled by 1005 * the generic interface code. If not defined ioctls return 1006 * not supported error code. 1007 * 1008 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map); 1009 * Used to set network devices bus interface parameters. This interface 1010 * is retained for legacy reasons; new devices should use the bus 1011 * interface (PCI) for low level management. 1012 * 1013 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu); 1014 * Called when a user wants to change the Maximum Transfer Unit 1015 * of a device. 1016 * 1017 * void (*ndo_tx_timeout)(struct net_device *dev); 1018 * Callback used when the transmitter has not made any progress 1019 * for dev->watchdog ticks. 1020 * 1021 * void (*ndo_get_stats64)(struct net_device *dev, 1022 * struct rtnl_link_stats64 *storage); 1023 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev); 1024 * Called when a user wants to get the network device usage 1025 * statistics. Drivers must do one of the following: 1026 * 1. Define @ndo_get_stats64 to fill in a zero-initialised 1027 * rtnl_link_stats64 structure passed by the caller. 1028 * 2. Define @ndo_get_stats to update a net_device_stats structure 1029 * (which should normally be dev->stats) and return a pointer to 1030 * it. The structure may be changed asynchronously only if each 1031 * field is written atomically. 1032 * 3. Update dev->stats asynchronously and atomically, and define 1033 * neither operation. 1034 * 1035 * bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id) 1036 * Return true if this device supports offload stats of this attr_id. 1037 * 1038 * int (*ndo_get_offload_stats)(int attr_id, const struct net_device *dev, 1039 * void *attr_data) 1040 * Get statistics for offload operations by attr_id. Write it into the 1041 * attr_data pointer. 1042 * 1043 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16 vid); 1044 * If device supports VLAN filtering this function is called when a 1045 * VLAN id is registered. 1046 * 1047 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, __be16 proto, u16 vid); 1048 * If device supports VLAN filtering this function is called when a 1049 * VLAN id is unregistered. 1050 * 1051 * void (*ndo_poll_controller)(struct net_device *dev); 1052 * 1053 * SR-IOV management functions. 1054 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac); 1055 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan, 1056 * u8 qos, __be16 proto); 1057 * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate, 1058 * int max_tx_rate); 1059 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting); 1060 * int (*ndo_set_vf_trust)(struct net_device *dev, int vf, bool setting); 1061 * int (*ndo_get_vf_config)(struct net_device *dev, 1062 * int vf, struct ifla_vf_info *ivf); 1063 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state); 1064 * int (*ndo_set_vf_port)(struct net_device *dev, int vf, 1065 * struct nlattr *port[]); 1066 * 1067 * Enable or disable the VF ability to query its RSS Redirection Table and 1068 * Hash Key. This is needed since on some devices VF share this information 1069 * with PF and querying it may introduce a theoretical security risk. 1070 * int (*ndo_set_vf_rss_query_en)(struct net_device *dev, int vf, bool setting); 1071 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb); 1072 * int (*ndo_setup_tc)(struct net_device *dev, enum tc_setup_type type, 1073 * void *type_data); 1074 * Called to setup any 'tc' scheduler, classifier or action on @dev. 1075 * This is always called from the stack with the rtnl lock held and netif 1076 * tx queues stopped. This allows the netdevice to perform queue 1077 * management safely. 1078 * 1079 * Fiber Channel over Ethernet (FCoE) offload functions. 1080 * int (*ndo_fcoe_enable)(struct net_device *dev); 1081 * Called when the FCoE protocol stack wants to start using LLD for FCoE 1082 * so the underlying device can perform whatever needed configuration or 1083 * initialization to support acceleration of FCoE traffic. 1084 * 1085 * int (*ndo_fcoe_disable)(struct net_device *dev); 1086 * Called when the FCoE protocol stack wants to stop using LLD for FCoE 1087 * so the underlying device can perform whatever needed clean-ups to 1088 * stop supporting acceleration of FCoE traffic. 1089 * 1090 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid, 1091 * struct scatterlist *sgl, unsigned int sgc); 1092 * Called when the FCoE Initiator wants to initialize an I/O that 1093 * is a possible candidate for Direct Data Placement (DDP). The LLD can 1094 * perform necessary setup and returns 1 to indicate the device is set up 1095 * successfully to perform DDP on this I/O, otherwise this returns 0. 1096 * 1097 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid); 1098 * Called when the FCoE Initiator/Target is done with the DDPed I/O as 1099 * indicated by the FC exchange id 'xid', so the underlying device can 1100 * clean up and reuse resources for later DDP requests. 1101 * 1102 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid, 1103 * struct scatterlist *sgl, unsigned int sgc); 1104 * Called when the FCoE Target wants to initialize an I/O that 1105 * is a possible candidate for Direct Data Placement (DDP). The LLD can 1106 * perform necessary setup and returns 1 to indicate the device is set up 1107 * successfully to perform DDP on this I/O, otherwise this returns 0. 1108 * 1109 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev, 1110 * struct netdev_fcoe_hbainfo *hbainfo); 1111 * Called when the FCoE Protocol stack wants information on the underlying 1112 * device. This information is utilized by the FCoE protocol stack to 1113 * register attributes with Fiber Channel management service as per the 1114 * FC-GS Fabric Device Management Information(FDMI) specification. 1115 * 1116 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type); 1117 * Called when the underlying device wants to override default World Wide 1118 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own 1119 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE 1120 * protocol stack to use. 1121 * 1122 * RFS acceleration. 1123 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb, 1124 * u16 rxq_index, u32 flow_id); 1125 * Set hardware filter for RFS. rxq_index is the target queue index; 1126 * flow_id is a flow ID to be passed to rps_may_expire_flow() later. 1127 * Return the filter ID on success, or a negative error code. 1128 * 1129 * Slave management functions (for bridge, bonding, etc). 1130 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev); 1131 * Called to make another netdev an underling. 1132 * 1133 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev); 1134 * Called to release previously enslaved netdev. 1135 * 1136 * Feature/offload setting functions. 1137 * netdev_features_t (*ndo_fix_features)(struct net_device *dev, 1138 * netdev_features_t features); 1139 * Adjusts the requested feature flags according to device-specific 1140 * constraints, and returns the resulting flags. Must not modify 1141 * the device state. 1142 * 1143 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features); 1144 * Called to update device configuration to new features. Passed 1145 * feature set might be less than what was returned by ndo_fix_features()). 1146 * Must return >0 or -errno if it changed dev->features itself. 1147 * 1148 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[], 1149 * struct net_device *dev, 1150 * const unsigned char *addr, u16 vid, u16 flags, 1151 * struct netlink_ext_ack *extack); 1152 * Adds an FDB entry to dev for addr. 1153 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[], 1154 * struct net_device *dev, 1155 * const unsigned char *addr, u16 vid) 1156 * Deletes the FDB entry from dev coresponding to addr. 1157 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb, 1158 * struct net_device *dev, struct net_device *filter_dev, 1159 * int *idx) 1160 * Used to add FDB entries to dump requests. Implementers should add 1161 * entries to skb and update idx with the number of entries. 1162 * 1163 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh, 1164 * u16 flags, struct netlink_ext_ack *extack) 1165 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq, 1166 * struct net_device *dev, u32 filter_mask, 1167 * int nlflags) 1168 * int (*ndo_bridge_dellink)(struct net_device *dev, struct nlmsghdr *nlh, 1169 * u16 flags); 1170 * 1171 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier); 1172 * Called to change device carrier. Soft-devices (like dummy, team, etc) 1173 * which do not represent real hardware may define this to allow their 1174 * userspace components to manage their virtual carrier state. Devices 1175 * that determine carrier state from physical hardware properties (eg 1176 * network cables) or protocol-dependent mechanisms (eg 1177 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function. 1178 * 1179 * int (*ndo_get_phys_port_id)(struct net_device *dev, 1180 * struct netdev_phys_item_id *ppid); 1181 * Called to get ID of physical port of this device. If driver does 1182 * not implement this, it is assumed that the hw is not able to have 1183 * multiple net devices on single physical port. 1184 * 1185 * int (*ndo_get_port_parent_id)(struct net_device *dev, 1186 * struct netdev_phys_item_id *ppid) 1187 * Called to get the parent ID of the physical port of this device. 1188 * 1189 * void (*ndo_udp_tunnel_add)(struct net_device *dev, 1190 * struct udp_tunnel_info *ti); 1191 * Called by UDP tunnel to notify a driver about the UDP port and socket 1192 * address family that a UDP tunnel is listnening to. It is called only 1193 * when a new port starts listening. The operation is protected by the 1194 * RTNL. 1195 * 1196 * void (*ndo_udp_tunnel_del)(struct net_device *dev, 1197 * struct udp_tunnel_info *ti); 1198 * Called by UDP tunnel to notify the driver about a UDP port and socket 1199 * address family that the UDP tunnel is not listening to anymore. The 1200 * operation is protected by the RTNL. 1201 * 1202 * void* (*ndo_dfwd_add_station)(struct net_device *pdev, 1203 * struct net_device *dev) 1204 * Called by upper layer devices to accelerate switching or other 1205 * station functionality into hardware. 'pdev is the lowerdev 1206 * to use for the offload and 'dev' is the net device that will 1207 * back the offload. Returns a pointer to the private structure 1208 * the upper layer will maintain. 1209 * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv) 1210 * Called by upper layer device to delete the station created 1211 * by 'ndo_dfwd_add_station'. 'pdev' is the net device backing 1212 * the station and priv is the structure returned by the add 1213 * operation. 1214 * int (*ndo_set_tx_maxrate)(struct net_device *dev, 1215 * int queue_index, u32 maxrate); 1216 * Called when a user wants to set a max-rate limitation of specific 1217 * TX queue. 1218 * int (*ndo_get_iflink)(const struct net_device *dev); 1219 * Called to get the iflink value of this device. 1220 * void (*ndo_change_proto_down)(struct net_device *dev, 1221 * bool proto_down); 1222 * This function is used to pass protocol port error state information 1223 * to the switch driver. The switch driver can react to the proto_down 1224 * by doing a phys down on the associated switch port. 1225 * int (*ndo_fill_metadata_dst)(struct net_device *dev, struct sk_buff *skb); 1226 * This function is used to get egress tunnel information for given skb. 1227 * This is useful for retrieving outer tunnel header parameters while 1228 * sampling packet. 1229 * void (*ndo_set_rx_headroom)(struct net_device *dev, int needed_headroom); 1230 * This function is used to specify the headroom that the skb must 1231 * consider when allocation skb during packet reception. Setting 1232 * appropriate rx headroom value allows avoiding skb head copy on 1233 * forward. Setting a negative value resets the rx headroom to the 1234 * default value. 1235 * int (*ndo_bpf)(struct net_device *dev, struct netdev_bpf *bpf); 1236 * This function is used to set or query state related to XDP on the 1237 * netdevice and manage BPF offload. See definition of 1238 * enum bpf_netdev_command for details. 1239 * int (*ndo_xdp_xmit)(struct net_device *dev, int n, struct xdp_frame **xdp, 1240 * u32 flags); 1241 * This function is used to submit @n XDP packets for transmit on a 1242 * netdevice. Returns number of frames successfully transmitted, frames 1243 * that got dropped are freed/returned via xdp_return_frame(). 1244 * Returns negative number, means general error invoking ndo, meaning 1245 * no frames were xmit'ed and core-caller will free all frames. 1246 * int (*ndo_xsk_wakeup)(struct net_device *dev, u32 queue_id, u32 flags); 1247 * This function is used to wake up the softirq, ksoftirqd or kthread 1248 * responsible for sending and/or receiving packets on a specific 1249 * queue id bound to an AF_XDP socket. The flags field specifies if 1250 * only RX, only Tx, or both should be woken up using the flags 1251 * XDP_WAKEUP_RX and XDP_WAKEUP_TX. 1252 * struct devlink_port *(*ndo_get_devlink_port)(struct net_device *dev); 1253 * Get devlink port instance associated with a given netdev. 1254 * Called with a reference on the netdevice and devlink locks only, 1255 * rtnl_lock is not held. 1256 */ 1257 struct net_device_ops { 1258 int (*ndo_init)(struct net_device *dev); 1259 void (*ndo_uninit)(struct net_device *dev); 1260 int (*ndo_open)(struct net_device *dev); 1261 int (*ndo_stop)(struct net_device *dev); 1262 netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb, 1263 struct net_device *dev); 1264 netdev_features_t (*ndo_features_check)(struct sk_buff *skb, 1265 struct net_device *dev, 1266 netdev_features_t features); 1267 u16 (*ndo_select_queue)(struct net_device *dev, 1268 struct sk_buff *skb, 1269 struct net_device *sb_dev); 1270 void (*ndo_change_rx_flags)(struct net_device *dev, 1271 int flags); 1272 void (*ndo_set_rx_mode)(struct net_device *dev); 1273 int (*ndo_set_mac_address)(struct net_device *dev, 1274 void *addr); 1275 int (*ndo_validate_addr)(struct net_device *dev); 1276 int (*ndo_do_ioctl)(struct net_device *dev, 1277 struct ifreq *ifr, int cmd); 1278 int (*ndo_set_config)(struct net_device *dev, 1279 struct ifmap *map); 1280 int (*ndo_change_mtu)(struct net_device *dev, 1281 int new_mtu); 1282 int (*ndo_neigh_setup)(struct net_device *dev, 1283 struct neigh_parms *); 1284 void (*ndo_tx_timeout) (struct net_device *dev); 1285 1286 void (*ndo_get_stats64)(struct net_device *dev, 1287 struct rtnl_link_stats64 *storage); 1288 bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id); 1289 int (*ndo_get_offload_stats)(int attr_id, 1290 const struct net_device *dev, 1291 void *attr_data); 1292 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev); 1293 1294 int (*ndo_vlan_rx_add_vid)(struct net_device *dev, 1295 __be16 proto, u16 vid); 1296 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, 1297 __be16 proto, u16 vid); 1298 #ifdef CONFIG_NET_POLL_CONTROLLER 1299 void (*ndo_poll_controller)(struct net_device *dev); 1300 int (*ndo_netpoll_setup)(struct net_device *dev, 1301 struct netpoll_info *info); 1302 void (*ndo_netpoll_cleanup)(struct net_device *dev); 1303 #endif 1304 int (*ndo_set_vf_mac)(struct net_device *dev, 1305 int queue, u8 *mac); 1306 int (*ndo_set_vf_vlan)(struct net_device *dev, 1307 int queue, u16 vlan, 1308 u8 qos, __be16 proto); 1309 int (*ndo_set_vf_rate)(struct net_device *dev, 1310 int vf, int min_tx_rate, 1311 int max_tx_rate); 1312 int (*ndo_set_vf_spoofchk)(struct net_device *dev, 1313 int vf, bool setting); 1314 int (*ndo_set_vf_trust)(struct net_device *dev, 1315 int vf, bool setting); 1316 int (*ndo_get_vf_config)(struct net_device *dev, 1317 int vf, 1318 struct ifla_vf_info *ivf); 1319 int (*ndo_set_vf_link_state)(struct net_device *dev, 1320 int vf, int link_state); 1321 int (*ndo_get_vf_stats)(struct net_device *dev, 1322 int vf, 1323 struct ifla_vf_stats 1324 *vf_stats); 1325 int (*ndo_set_vf_port)(struct net_device *dev, 1326 int vf, 1327 struct nlattr *port[]); 1328 int (*ndo_get_vf_port)(struct net_device *dev, 1329 int vf, struct sk_buff *skb); 1330 int (*ndo_get_vf_guid)(struct net_device *dev, 1331 int vf, 1332 struct ifla_vf_guid *node_guid, 1333 struct ifla_vf_guid *port_guid); 1334 int (*ndo_set_vf_guid)(struct net_device *dev, 1335 int vf, u64 guid, 1336 int guid_type); 1337 int (*ndo_set_vf_rss_query_en)( 1338 struct net_device *dev, 1339 int vf, bool setting); 1340 int (*ndo_setup_tc)(struct net_device *dev, 1341 enum tc_setup_type type, 1342 void *type_data); 1343 #if IS_ENABLED(CONFIG_FCOE) 1344 int (*ndo_fcoe_enable)(struct net_device *dev); 1345 int (*ndo_fcoe_disable)(struct net_device *dev); 1346 int (*ndo_fcoe_ddp_setup)(struct net_device *dev, 1347 u16 xid, 1348 struct scatterlist *sgl, 1349 unsigned int sgc); 1350 int (*ndo_fcoe_ddp_done)(struct net_device *dev, 1351 u16 xid); 1352 int (*ndo_fcoe_ddp_target)(struct net_device *dev, 1353 u16 xid, 1354 struct scatterlist *sgl, 1355 unsigned int sgc); 1356 int (*ndo_fcoe_get_hbainfo)(struct net_device *dev, 1357 struct netdev_fcoe_hbainfo *hbainfo); 1358 #endif 1359 1360 #if IS_ENABLED(CONFIG_LIBFCOE) 1361 #define NETDEV_FCOE_WWNN 0 1362 #define NETDEV_FCOE_WWPN 1 1363 int (*ndo_fcoe_get_wwn)(struct net_device *dev, 1364 u64 *wwn, int type); 1365 #endif 1366 1367 #ifdef CONFIG_RFS_ACCEL 1368 int (*ndo_rx_flow_steer)(struct net_device *dev, 1369 const struct sk_buff *skb, 1370 u16 rxq_index, 1371 u32 flow_id); 1372 #endif 1373 int (*ndo_add_slave)(struct net_device *dev, 1374 struct net_device *slave_dev, 1375 struct netlink_ext_ack *extack); 1376 int (*ndo_del_slave)(struct net_device *dev, 1377 struct net_device *slave_dev); 1378 netdev_features_t (*ndo_fix_features)(struct net_device *dev, 1379 netdev_features_t features); 1380 int (*ndo_set_features)(struct net_device *dev, 1381 netdev_features_t features); 1382 int (*ndo_neigh_construct)(struct net_device *dev, 1383 struct neighbour *n); 1384 void (*ndo_neigh_destroy)(struct net_device *dev, 1385 struct neighbour *n); 1386 1387 int (*ndo_fdb_add)(struct ndmsg *ndm, 1388 struct nlattr *tb[], 1389 struct net_device *dev, 1390 const unsigned char *addr, 1391 u16 vid, 1392 u16 flags, 1393 struct netlink_ext_ack *extack); 1394 int (*ndo_fdb_del)(struct ndmsg *ndm, 1395 struct nlattr *tb[], 1396 struct net_device *dev, 1397 const unsigned char *addr, 1398 u16 vid); 1399 int (*ndo_fdb_dump)(struct sk_buff *skb, 1400 struct netlink_callback *cb, 1401 struct net_device *dev, 1402 struct net_device *filter_dev, 1403 int *idx); 1404 int (*ndo_fdb_get)(struct sk_buff *skb, 1405 struct nlattr *tb[], 1406 struct net_device *dev, 1407 const unsigned char *addr, 1408 u16 vid, u32 portid, u32 seq, 1409 struct netlink_ext_ack *extack); 1410 int (*ndo_bridge_setlink)(struct net_device *dev, 1411 struct nlmsghdr *nlh, 1412 u16 flags, 1413 struct netlink_ext_ack *extack); 1414 int (*ndo_bridge_getlink)(struct sk_buff *skb, 1415 u32 pid, u32 seq, 1416 struct net_device *dev, 1417 u32 filter_mask, 1418 int nlflags); 1419 int (*ndo_bridge_dellink)(struct net_device *dev, 1420 struct nlmsghdr *nlh, 1421 u16 flags); 1422 int (*ndo_change_carrier)(struct net_device *dev, 1423 bool new_carrier); 1424 int (*ndo_get_phys_port_id)(struct net_device *dev, 1425 struct netdev_phys_item_id *ppid); 1426 int (*ndo_get_port_parent_id)(struct net_device *dev, 1427 struct netdev_phys_item_id *ppid); 1428 int (*ndo_get_phys_port_name)(struct net_device *dev, 1429 char *name, size_t len); 1430 void (*ndo_udp_tunnel_add)(struct net_device *dev, 1431 struct udp_tunnel_info *ti); 1432 void (*ndo_udp_tunnel_del)(struct net_device *dev, 1433 struct udp_tunnel_info *ti); 1434 void* (*ndo_dfwd_add_station)(struct net_device *pdev, 1435 struct net_device *dev); 1436 void (*ndo_dfwd_del_station)(struct net_device *pdev, 1437 void *priv); 1438 1439 int (*ndo_set_tx_maxrate)(struct net_device *dev, 1440 int queue_index, 1441 u32 maxrate); 1442 int (*ndo_get_iflink)(const struct net_device *dev); 1443 int (*ndo_change_proto_down)(struct net_device *dev, 1444 bool proto_down); 1445 int (*ndo_fill_metadata_dst)(struct net_device *dev, 1446 struct sk_buff *skb); 1447 void (*ndo_set_rx_headroom)(struct net_device *dev, 1448 int needed_headroom); 1449 int (*ndo_bpf)(struct net_device *dev, 1450 struct netdev_bpf *bpf); 1451 int (*ndo_xdp_xmit)(struct net_device *dev, int n, 1452 struct xdp_frame **xdp, 1453 u32 flags); 1454 int (*ndo_xsk_wakeup)(struct net_device *dev, 1455 u32 queue_id, u32 flags); 1456 struct devlink_port * (*ndo_get_devlink_port)(struct net_device *dev); 1457 }; 1458 1459 /** 1460 * enum net_device_priv_flags - &struct net_device priv_flags 1461 * 1462 * These are the &struct net_device, they are only set internally 1463 * by drivers and used in the kernel. These flags are invisible to 1464 * userspace; this means that the order of these flags can change 1465 * during any kernel release. 1466 * 1467 * You should have a pretty good reason to be extending these flags. 1468 * 1469 * @IFF_802_1Q_VLAN: 802.1Q VLAN device 1470 * @IFF_EBRIDGE: Ethernet bridging device 1471 * @IFF_BONDING: bonding master or slave 1472 * @IFF_ISATAP: ISATAP interface (RFC4214) 1473 * @IFF_WAN_HDLC: WAN HDLC device 1474 * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to 1475 * release skb->dst 1476 * @IFF_DONT_BRIDGE: disallow bridging this ether dev 1477 * @IFF_DISABLE_NETPOLL: disable netpoll at run-time 1478 * @IFF_MACVLAN_PORT: device used as macvlan port 1479 * @IFF_BRIDGE_PORT: device used as bridge port 1480 * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port 1481 * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit 1482 * @IFF_UNICAST_FLT: Supports unicast filtering 1483 * @IFF_TEAM_PORT: device used as team port 1484 * @IFF_SUPP_NOFCS: device supports sending custom FCS 1485 * @IFF_LIVE_ADDR_CHANGE: device supports hardware address 1486 * change when it's running 1487 * @IFF_MACVLAN: Macvlan device 1488 * @IFF_XMIT_DST_RELEASE_PERM: IFF_XMIT_DST_RELEASE not taking into account 1489 * underlying stacked devices 1490 * @IFF_L3MDEV_MASTER: device is an L3 master device 1491 * @IFF_NO_QUEUE: device can run without qdisc attached 1492 * @IFF_OPENVSWITCH: device is a Open vSwitch master 1493 * @IFF_L3MDEV_SLAVE: device is enslaved to an L3 master device 1494 * @IFF_TEAM: device is a team device 1495 * @IFF_RXFH_CONFIGURED: device has had Rx Flow indirection table configured 1496 * @IFF_PHONY_HEADROOM: the headroom value is controlled by an external 1497 * entity (i.e. the master device for bridged veth) 1498 * @IFF_MACSEC: device is a MACsec device 1499 * @IFF_NO_RX_HANDLER: device doesn't support the rx_handler hook 1500 * @IFF_FAILOVER: device is a failover master device 1501 * @IFF_FAILOVER_SLAVE: device is lower dev of a failover master device 1502 * @IFF_L3MDEV_RX_HANDLER: only invoke the rx handler of L3 master device 1503 * @IFF_LIVE_RENAME_OK: rename is allowed while device is up and running 1504 */ 1505 enum netdev_priv_flags { 1506 IFF_802_1Q_VLAN = 1<<0, 1507 IFF_EBRIDGE = 1<<1, 1508 IFF_BONDING = 1<<2, 1509 IFF_ISATAP = 1<<3, 1510 IFF_WAN_HDLC = 1<<4, 1511 IFF_XMIT_DST_RELEASE = 1<<5, 1512 IFF_DONT_BRIDGE = 1<<6, 1513 IFF_DISABLE_NETPOLL = 1<<7, 1514 IFF_MACVLAN_PORT = 1<<8, 1515 IFF_BRIDGE_PORT = 1<<9, 1516 IFF_OVS_DATAPATH = 1<<10, 1517 IFF_TX_SKB_SHARING = 1<<11, 1518 IFF_UNICAST_FLT = 1<<12, 1519 IFF_TEAM_PORT = 1<<13, 1520 IFF_SUPP_NOFCS = 1<<14, 1521 IFF_LIVE_ADDR_CHANGE = 1<<15, 1522 IFF_MACVLAN = 1<<16, 1523 IFF_XMIT_DST_RELEASE_PERM = 1<<17, 1524 IFF_L3MDEV_MASTER = 1<<18, 1525 IFF_NO_QUEUE = 1<<19, 1526 IFF_OPENVSWITCH = 1<<20, 1527 IFF_L3MDEV_SLAVE = 1<<21, 1528 IFF_TEAM = 1<<22, 1529 IFF_RXFH_CONFIGURED = 1<<23, 1530 IFF_PHONY_HEADROOM = 1<<24, 1531 IFF_MACSEC = 1<<25, 1532 IFF_NO_RX_HANDLER = 1<<26, 1533 IFF_FAILOVER = 1<<27, 1534 IFF_FAILOVER_SLAVE = 1<<28, 1535 IFF_L3MDEV_RX_HANDLER = 1<<29, 1536 IFF_LIVE_RENAME_OK = 1<<30, 1537 }; 1538 1539 #define IFF_802_1Q_VLAN IFF_802_1Q_VLAN 1540 #define IFF_EBRIDGE IFF_EBRIDGE 1541 #define IFF_BONDING IFF_BONDING 1542 #define IFF_ISATAP IFF_ISATAP 1543 #define IFF_WAN_HDLC IFF_WAN_HDLC 1544 #define IFF_XMIT_DST_RELEASE IFF_XMIT_DST_RELEASE 1545 #define IFF_DONT_BRIDGE IFF_DONT_BRIDGE 1546 #define IFF_DISABLE_NETPOLL IFF_DISABLE_NETPOLL 1547 #define IFF_MACVLAN_PORT IFF_MACVLAN_PORT 1548 #define IFF_BRIDGE_PORT IFF_BRIDGE_PORT 1549 #define IFF_OVS_DATAPATH IFF_OVS_DATAPATH 1550 #define IFF_TX_SKB_SHARING IFF_TX_SKB_SHARING 1551 #define IFF_UNICAST_FLT IFF_UNICAST_FLT 1552 #define IFF_TEAM_PORT IFF_TEAM_PORT 1553 #define IFF_SUPP_NOFCS IFF_SUPP_NOFCS 1554 #define IFF_LIVE_ADDR_CHANGE IFF_LIVE_ADDR_CHANGE 1555 #define IFF_MACVLAN IFF_MACVLAN 1556 #define IFF_XMIT_DST_RELEASE_PERM IFF_XMIT_DST_RELEASE_PERM 1557 #define IFF_L3MDEV_MASTER IFF_L3MDEV_MASTER 1558 #define IFF_NO_QUEUE IFF_NO_QUEUE 1559 #define IFF_OPENVSWITCH IFF_OPENVSWITCH 1560 #define IFF_L3MDEV_SLAVE IFF_L3MDEV_SLAVE 1561 #define IFF_TEAM IFF_TEAM 1562 #define IFF_RXFH_CONFIGURED IFF_RXFH_CONFIGURED 1563 #define IFF_MACSEC IFF_MACSEC 1564 #define IFF_NO_RX_HANDLER IFF_NO_RX_HANDLER 1565 #define IFF_FAILOVER IFF_FAILOVER 1566 #define IFF_FAILOVER_SLAVE IFF_FAILOVER_SLAVE 1567 #define IFF_L3MDEV_RX_HANDLER IFF_L3MDEV_RX_HANDLER 1568 #define IFF_LIVE_RENAME_OK IFF_LIVE_RENAME_OK 1569 1570 /** 1571 * struct net_device - The DEVICE structure. 1572 * 1573 * Actually, this whole structure is a big mistake. It mixes I/O 1574 * data with strictly "high-level" data, and it has to know about 1575 * almost every data structure used in the INET module. 1576 * 1577 * @name: This is the first field of the "visible" part of this structure 1578 * (i.e. as seen by users in the "Space.c" file). It is the name 1579 * of the interface. 1580 * 1581 * @name_node: Name hashlist node 1582 * @ifalias: SNMP alias 1583 * @mem_end: Shared memory end 1584 * @mem_start: Shared memory start 1585 * @base_addr: Device I/O address 1586 * @irq: Device IRQ number 1587 * 1588 * @state: Generic network queuing layer state, see netdev_state_t 1589 * @dev_list: The global list of network devices 1590 * @napi_list: List entry used for polling NAPI devices 1591 * @unreg_list: List entry when we are unregistering the 1592 * device; see the function unregister_netdev 1593 * @close_list: List entry used when we are closing the device 1594 * @ptype_all: Device-specific packet handlers for all protocols 1595 * @ptype_specific: Device-specific, protocol-specific packet handlers 1596 * 1597 * @adj_list: Directly linked devices, like slaves for bonding 1598 * @features: Currently active device features 1599 * @hw_features: User-changeable features 1600 * 1601 * @wanted_features: User-requested features 1602 * @vlan_features: Mask of features inheritable by VLAN devices 1603 * 1604 * @hw_enc_features: Mask of features inherited by encapsulating devices 1605 * This field indicates what encapsulation 1606 * offloads the hardware is capable of doing, 1607 * and drivers will need to set them appropriately. 1608 * 1609 * @mpls_features: Mask of features inheritable by MPLS 1610 * 1611 * @ifindex: interface index 1612 * @group: The group the device belongs to 1613 * 1614 * @stats: Statistics struct, which was left as a legacy, use 1615 * rtnl_link_stats64 instead 1616 * 1617 * @rx_dropped: Dropped packets by core network, 1618 * do not use this in drivers 1619 * @tx_dropped: Dropped packets by core network, 1620 * do not use this in drivers 1621 * @rx_nohandler: nohandler dropped packets by core network on 1622 * inactive devices, do not use this in drivers 1623 * @carrier_up_count: Number of times the carrier has been up 1624 * @carrier_down_count: Number of times the carrier has been down 1625 * 1626 * @wireless_handlers: List of functions to handle Wireless Extensions, 1627 * instead of ioctl, 1628 * see <net/iw_handler.h> for details. 1629 * @wireless_data: Instance data managed by the core of wireless extensions 1630 * 1631 * @netdev_ops: Includes several pointers to callbacks, 1632 * if one wants to override the ndo_*() functions 1633 * @ethtool_ops: Management operations 1634 * @ndisc_ops: Includes callbacks for different IPv6 neighbour 1635 * discovery handling. Necessary for e.g. 6LoWPAN. 1636 * @header_ops: Includes callbacks for creating,parsing,caching,etc 1637 * of Layer 2 headers. 1638 * 1639 * @flags: Interface flags (a la BSD) 1640 * @priv_flags: Like 'flags' but invisible to userspace, 1641 * see if.h for the definitions 1642 * @gflags: Global flags ( kept as legacy ) 1643 * @padded: How much padding added by alloc_netdev() 1644 * @operstate: RFC2863 operstate 1645 * @link_mode: Mapping policy to operstate 1646 * @if_port: Selectable AUI, TP, ... 1647 * @dma: DMA channel 1648 * @mtu: Interface MTU value 1649 * @min_mtu: Interface Minimum MTU value 1650 * @max_mtu: Interface Maximum MTU value 1651 * @type: Interface hardware type 1652 * @hard_header_len: Maximum hardware header length. 1653 * @min_header_len: Minimum hardware header length 1654 * 1655 * @needed_headroom: Extra headroom the hardware may need, but not in all 1656 * cases can this be guaranteed 1657 * @needed_tailroom: Extra tailroom the hardware may need, but not in all 1658 * cases can this be guaranteed. Some cases also use 1659 * LL_MAX_HEADER instead to allocate the skb 1660 * 1661 * interface address info: 1662 * 1663 * @perm_addr: Permanent hw address 1664 * @addr_assign_type: Hw address assignment type 1665 * @addr_len: Hardware address length 1666 * @upper_level: Maximum depth level of upper devices. 1667 * @lower_level: Maximum depth level of lower devices. 1668 * @neigh_priv_len: Used in neigh_alloc() 1669 * @dev_id: Used to differentiate devices that share 1670 * the same link layer address 1671 * @dev_port: Used to differentiate devices that share 1672 * the same function 1673 * @addr_list_lock: XXX: need comments on this one 1674 * @uc_promisc: Counter that indicates promiscuous mode 1675 * has been enabled due to the need to listen to 1676 * additional unicast addresses in a device that 1677 * does not implement ndo_set_rx_mode() 1678 * @uc: unicast mac addresses 1679 * @mc: multicast mac addresses 1680 * @dev_addrs: list of device hw addresses 1681 * @queues_kset: Group of all Kobjects in the Tx and RX queues 1682 * @promiscuity: Number of times the NIC is told to work in 1683 * promiscuous mode; if it becomes 0 the NIC will 1684 * exit promiscuous mode 1685 * @allmulti: Counter, enables or disables allmulticast mode 1686 * 1687 * @vlan_info: VLAN info 1688 * @dsa_ptr: dsa specific data 1689 * @tipc_ptr: TIPC specific data 1690 * @atalk_ptr: AppleTalk link 1691 * @ip_ptr: IPv4 specific data 1692 * @dn_ptr: DECnet specific data 1693 * @ip6_ptr: IPv6 specific data 1694 * @ax25_ptr: AX.25 specific data 1695 * @ieee80211_ptr: IEEE 802.11 specific data, assign before registering 1696 * 1697 * @dev_addr: Hw address (before bcast, 1698 * because most packets are unicast) 1699 * 1700 * @_rx: Array of RX queues 1701 * @num_rx_queues: Number of RX queues 1702 * allocated at register_netdev() time 1703 * @real_num_rx_queues: Number of RX queues currently active in device 1704 * 1705 * @rx_handler: handler for received packets 1706 * @rx_handler_data: XXX: need comments on this one 1707 * @miniq_ingress: ingress/clsact qdisc specific data for 1708 * ingress processing 1709 * @ingress_queue: XXX: need comments on this one 1710 * @broadcast: hw bcast address 1711 * 1712 * @rx_cpu_rmap: CPU reverse-mapping for RX completion interrupts, 1713 * indexed by RX queue number. Assigned by driver. 1714 * This must only be set if the ndo_rx_flow_steer 1715 * operation is defined 1716 * @index_hlist: Device index hash chain 1717 * 1718 * @_tx: Array of TX queues 1719 * @num_tx_queues: Number of TX queues allocated at alloc_netdev_mq() time 1720 * @real_num_tx_queues: Number of TX queues currently active in device 1721 * @qdisc: Root qdisc from userspace point of view 1722 * @tx_queue_len: Max frames per queue allowed 1723 * @tx_global_lock: XXX: need comments on this one 1724 * 1725 * @xps_maps: XXX: need comments on this one 1726 * @miniq_egress: clsact qdisc specific data for 1727 * egress processing 1728 * @watchdog_timeo: Represents the timeout that is used by 1729 * the watchdog (see dev_watchdog()) 1730 * @watchdog_timer: List of timers 1731 * 1732 * @pcpu_refcnt: Number of references to this device 1733 * @todo_list: Delayed register/unregister 1734 * @link_watch_list: XXX: need comments on this one 1735 * 1736 * @reg_state: Register/unregister state machine 1737 * @dismantle: Device is going to be freed 1738 * @rtnl_link_state: This enum represents the phases of creating 1739 * a new link 1740 * 1741 * @needs_free_netdev: Should unregister perform free_netdev? 1742 * @priv_destructor: Called from unregister 1743 * @npinfo: XXX: need comments on this one 1744 * @nd_net: Network namespace this network device is inside 1745 * 1746 * @ml_priv: Mid-layer private 1747 * @lstats: Loopback statistics 1748 * @tstats: Tunnel statistics 1749 * @dstats: Dummy statistics 1750 * @vstats: Virtual ethernet statistics 1751 * 1752 * @garp_port: GARP 1753 * @mrp_port: MRP 1754 * 1755 * @dev: Class/net/name entry 1756 * @sysfs_groups: Space for optional device, statistics and wireless 1757 * sysfs groups 1758 * 1759 * @sysfs_rx_queue_group: Space for optional per-rx queue attributes 1760 * @rtnl_link_ops: Rtnl_link_ops 1761 * 1762 * @gso_max_size: Maximum size of generic segmentation offload 1763 * @gso_max_segs: Maximum number of segments that can be passed to the 1764 * NIC for GSO 1765 * 1766 * @dcbnl_ops: Data Center Bridging netlink ops 1767 * @num_tc: Number of traffic classes in the net device 1768 * @tc_to_txq: XXX: need comments on this one 1769 * @prio_tc_map: XXX: need comments on this one 1770 * 1771 * @fcoe_ddp_xid: Max exchange id for FCoE LRO by ddp 1772 * 1773 * @priomap: XXX: need comments on this one 1774 * @phydev: Physical device may attach itself 1775 * for hardware timestamping 1776 * @sfp_bus: attached &struct sfp_bus structure. 1777 * @qdisc_tx_busylock_key: lockdep class annotating Qdisc->busylock 1778 spinlock 1779 * @qdisc_running_key: lockdep class annotating Qdisc->running seqcount 1780 * @qdisc_xmit_lock_key: lockdep class annotating 1781 * netdev_queue->_xmit_lock spinlock 1782 * @addr_list_lock_key: lockdep class annotating 1783 * net_device->addr_list_lock spinlock 1784 * 1785 * @proto_down: protocol port state information can be sent to the 1786 * switch driver and used to set the phys state of the 1787 * switch port. 1788 * 1789 * @wol_enabled: Wake-on-LAN is enabled 1790 * 1791 * FIXME: cleanup struct net_device such that network protocol info 1792 * moves out. 1793 */ 1794 1795 struct net_device { 1796 char name[IFNAMSIZ]; 1797 struct netdev_name_node *name_node; 1798 struct dev_ifalias __rcu *ifalias; 1799 /* 1800 * I/O specific fields 1801 * FIXME: Merge these and struct ifmap into one 1802 */ 1803 unsigned long mem_end; 1804 unsigned long mem_start; 1805 unsigned long base_addr; 1806 int irq; 1807 1808 /* 1809 * Some hardware also needs these fields (state,dev_list, 1810 * napi_list,unreg_list,close_list) but they are not 1811 * part of the usual set specified in Space.c. 1812 */ 1813 1814 unsigned long state; 1815 1816 struct list_head dev_list; 1817 struct list_head napi_list; 1818 struct list_head unreg_list; 1819 struct list_head close_list; 1820 struct list_head ptype_all; 1821 struct list_head ptype_specific; 1822 1823 struct { 1824 struct list_head upper; 1825 struct list_head lower; 1826 } adj_list; 1827 1828 netdev_features_t features; 1829 netdev_features_t hw_features; 1830 netdev_features_t wanted_features; 1831 netdev_features_t vlan_features; 1832 netdev_features_t hw_enc_features; 1833 netdev_features_t mpls_features; 1834 netdev_features_t gso_partial_features; 1835 1836 int ifindex; 1837 int group; 1838 1839 struct net_device_stats stats; 1840 1841 atomic_long_t rx_dropped; 1842 atomic_long_t tx_dropped; 1843 atomic_long_t rx_nohandler; 1844 1845 /* Stats to monitor link on/off, flapping */ 1846 atomic_t carrier_up_count; 1847 atomic_t carrier_down_count; 1848 1849 #ifdef CONFIG_WIRELESS_EXT 1850 const struct iw_handler_def *wireless_handlers; 1851 struct iw_public_data *wireless_data; 1852 #endif 1853 const struct net_device_ops *netdev_ops; 1854 const struct ethtool_ops *ethtool_ops; 1855 #ifdef CONFIG_NET_L3_MASTER_DEV 1856 const struct l3mdev_ops *l3mdev_ops; 1857 #endif 1858 #if IS_ENABLED(CONFIG_IPV6) 1859 const struct ndisc_ops *ndisc_ops; 1860 #endif 1861 1862 #ifdef CONFIG_XFRM_OFFLOAD 1863 const struct xfrmdev_ops *xfrmdev_ops; 1864 #endif 1865 1866 #if IS_ENABLED(CONFIG_TLS_DEVICE) 1867 const struct tlsdev_ops *tlsdev_ops; 1868 #endif 1869 1870 const struct header_ops *header_ops; 1871 1872 unsigned int flags; 1873 unsigned int priv_flags; 1874 1875 unsigned short gflags; 1876 unsigned short padded; 1877 1878 unsigned char operstate; 1879 unsigned char link_mode; 1880 1881 unsigned char if_port; 1882 unsigned char dma; 1883 1884 unsigned int mtu; 1885 unsigned int min_mtu; 1886 unsigned int max_mtu; 1887 unsigned short type; 1888 unsigned short hard_header_len; 1889 unsigned char min_header_len; 1890 1891 unsigned short needed_headroom; 1892 unsigned short needed_tailroom; 1893 1894 /* Interface address info. */ 1895 unsigned char perm_addr[MAX_ADDR_LEN]; 1896 unsigned char addr_assign_type; 1897 unsigned char addr_len; 1898 unsigned char upper_level; 1899 unsigned char lower_level; 1900 unsigned short neigh_priv_len; 1901 unsigned short dev_id; 1902 unsigned short dev_port; 1903 spinlock_t addr_list_lock; 1904 unsigned char name_assign_type; 1905 bool uc_promisc; 1906 struct netdev_hw_addr_list uc; 1907 struct netdev_hw_addr_list mc; 1908 struct netdev_hw_addr_list dev_addrs; 1909 1910 #ifdef CONFIG_SYSFS 1911 struct kset *queues_kset; 1912 #endif 1913 unsigned int promiscuity; 1914 unsigned int allmulti; 1915 1916 1917 /* Protocol-specific pointers */ 1918 1919 #if IS_ENABLED(CONFIG_VLAN_8021Q) 1920 struct vlan_info __rcu *vlan_info; 1921 #endif 1922 #if IS_ENABLED(CONFIG_NET_DSA) 1923 struct dsa_port *dsa_ptr; 1924 #endif 1925 #if IS_ENABLED(CONFIG_TIPC) 1926 struct tipc_bearer __rcu *tipc_ptr; 1927 #endif 1928 #if IS_ENABLED(CONFIG_IRDA) || IS_ENABLED(CONFIG_ATALK) 1929 void *atalk_ptr; 1930 #endif 1931 struct in_device __rcu *ip_ptr; 1932 #if IS_ENABLED(CONFIG_DECNET) 1933 struct dn_dev __rcu *dn_ptr; 1934 #endif 1935 struct inet6_dev __rcu *ip6_ptr; 1936 #if IS_ENABLED(CONFIG_AX25) 1937 void *ax25_ptr; 1938 #endif 1939 struct wireless_dev *ieee80211_ptr; 1940 struct wpan_dev *ieee802154_ptr; 1941 #if IS_ENABLED(CONFIG_MPLS_ROUTING) 1942 struct mpls_dev __rcu *mpls_ptr; 1943 #endif 1944 1945 /* 1946 * Cache lines mostly used on receive path (including eth_type_trans()) 1947 */ 1948 /* Interface address info used in eth_type_trans() */ 1949 unsigned char *dev_addr; 1950 1951 struct netdev_rx_queue *_rx; 1952 unsigned int num_rx_queues; 1953 unsigned int real_num_rx_queues; 1954 1955 struct bpf_prog __rcu *xdp_prog; 1956 unsigned long gro_flush_timeout; 1957 rx_handler_func_t __rcu *rx_handler; 1958 void __rcu *rx_handler_data; 1959 1960 #ifdef CONFIG_NET_CLS_ACT 1961 struct mini_Qdisc __rcu *miniq_ingress; 1962 #endif 1963 struct netdev_queue __rcu *ingress_queue; 1964 #ifdef CONFIG_NETFILTER_INGRESS 1965 struct nf_hook_entries __rcu *nf_hooks_ingress; 1966 #endif 1967 1968 unsigned char broadcast[MAX_ADDR_LEN]; 1969 #ifdef CONFIG_RFS_ACCEL 1970 struct cpu_rmap *rx_cpu_rmap; 1971 #endif 1972 struct hlist_node index_hlist; 1973 1974 /* 1975 * Cache lines mostly used on transmit path 1976 */ 1977 struct netdev_queue *_tx ____cacheline_aligned_in_smp; 1978 unsigned int num_tx_queues; 1979 unsigned int real_num_tx_queues; 1980 struct Qdisc *qdisc; 1981 #ifdef CONFIG_NET_SCHED 1982 DECLARE_HASHTABLE (qdisc_hash, 4); 1983 #endif 1984 unsigned int tx_queue_len; 1985 spinlock_t tx_global_lock; 1986 int watchdog_timeo; 1987 1988 #ifdef CONFIG_XPS 1989 struct xps_dev_maps __rcu *xps_cpus_map; 1990 struct xps_dev_maps __rcu *xps_rxqs_map; 1991 #endif 1992 #ifdef CONFIG_NET_CLS_ACT 1993 struct mini_Qdisc __rcu *miniq_egress; 1994 #endif 1995 1996 /* These may be needed for future network-power-down code. */ 1997 struct timer_list watchdog_timer; 1998 1999 int __percpu *pcpu_refcnt; 2000 struct list_head todo_list; 2001 2002 struct list_head link_watch_list; 2003 2004 enum { NETREG_UNINITIALIZED=0, 2005 NETREG_REGISTERED, /* completed register_netdevice */ 2006 NETREG_UNREGISTERING, /* called unregister_netdevice */ 2007 NETREG_UNREGISTERED, /* completed unregister todo */ 2008 NETREG_RELEASED, /* called free_netdev */ 2009 NETREG_DUMMY, /* dummy device for NAPI poll */ 2010 } reg_state:8; 2011 2012 bool dismantle; 2013 2014 enum { 2015 RTNL_LINK_INITIALIZED, 2016 RTNL_LINK_INITIALIZING, 2017 } rtnl_link_state:16; 2018 2019 bool needs_free_netdev; 2020 void (*priv_destructor)(struct net_device *dev); 2021 2022 #ifdef CONFIG_NETPOLL 2023 struct netpoll_info __rcu *npinfo; 2024 #endif 2025 2026 possible_net_t nd_net; 2027 2028 /* mid-layer private */ 2029 union { 2030 void *ml_priv; 2031 struct pcpu_lstats __percpu *lstats; 2032 struct pcpu_sw_netstats __percpu *tstats; 2033 struct pcpu_dstats __percpu *dstats; 2034 }; 2035 2036 #if IS_ENABLED(CONFIG_GARP) 2037 struct garp_port __rcu *garp_port; 2038 #endif 2039 #if IS_ENABLED(CONFIG_MRP) 2040 struct mrp_port __rcu *mrp_port; 2041 #endif 2042 2043 struct device dev; 2044 const struct attribute_group *sysfs_groups[4]; 2045 const struct attribute_group *sysfs_rx_queue_group; 2046 2047 const struct rtnl_link_ops *rtnl_link_ops; 2048 2049 /* for setting kernel sock attribute on TCP connection setup */ 2050 #define GSO_MAX_SIZE 65536 2051 unsigned int gso_max_size; 2052 #define GSO_MAX_SEGS 65535 2053 u16 gso_max_segs; 2054 2055 #ifdef CONFIG_DCB 2056 const struct dcbnl_rtnl_ops *dcbnl_ops; 2057 #endif 2058 s16 num_tc; 2059 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE]; 2060 u8 prio_tc_map[TC_BITMASK + 1]; 2061 2062 #if IS_ENABLED(CONFIG_FCOE) 2063 unsigned int fcoe_ddp_xid; 2064 #endif 2065 #if IS_ENABLED(CONFIG_CGROUP_NET_PRIO) 2066 struct netprio_map __rcu *priomap; 2067 #endif 2068 struct phy_device *phydev; 2069 struct sfp_bus *sfp_bus; 2070 struct lock_class_key qdisc_tx_busylock_key; 2071 struct lock_class_key qdisc_running_key; 2072 struct lock_class_key qdisc_xmit_lock_key; 2073 struct lock_class_key addr_list_lock_key; 2074 bool proto_down; 2075 unsigned wol_enabled:1; 2076 }; 2077 #define to_net_dev(d) container_of(d, struct net_device, dev) 2078 2079 static inline bool netif_elide_gro(const struct net_device *dev) 2080 { 2081 if (!(dev->features & NETIF_F_GRO) || dev->xdp_prog) 2082 return true; 2083 return false; 2084 } 2085 2086 #define NETDEV_ALIGN 32 2087 2088 static inline 2089 int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio) 2090 { 2091 return dev->prio_tc_map[prio & TC_BITMASK]; 2092 } 2093 2094 static inline 2095 int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc) 2096 { 2097 if (tc >= dev->num_tc) 2098 return -EINVAL; 2099 2100 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK; 2101 return 0; 2102 } 2103 2104 int netdev_txq_to_tc(struct net_device *dev, unsigned int txq); 2105 void netdev_reset_tc(struct net_device *dev); 2106 int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset); 2107 int netdev_set_num_tc(struct net_device *dev, u8 num_tc); 2108 2109 static inline 2110 int netdev_get_num_tc(struct net_device *dev) 2111 { 2112 return dev->num_tc; 2113 } 2114 2115 void netdev_unbind_sb_channel(struct net_device *dev, 2116 struct net_device *sb_dev); 2117 int netdev_bind_sb_channel_queue(struct net_device *dev, 2118 struct net_device *sb_dev, 2119 u8 tc, u16 count, u16 offset); 2120 int netdev_set_sb_channel(struct net_device *dev, u16 channel); 2121 static inline int netdev_get_sb_channel(struct net_device *dev) 2122 { 2123 return max_t(int, -dev->num_tc, 0); 2124 } 2125 2126 static inline 2127 struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev, 2128 unsigned int index) 2129 { 2130 return &dev->_tx[index]; 2131 } 2132 2133 static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev, 2134 const struct sk_buff *skb) 2135 { 2136 return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb)); 2137 } 2138 2139 static inline void netdev_for_each_tx_queue(struct net_device *dev, 2140 void (*f)(struct net_device *, 2141 struct netdev_queue *, 2142 void *), 2143 void *arg) 2144 { 2145 unsigned int i; 2146 2147 for (i = 0; i < dev->num_tx_queues; i++) 2148 f(dev, &dev->_tx[i], arg); 2149 } 2150 2151 u16 netdev_pick_tx(struct net_device *dev, struct sk_buff *skb, 2152 struct net_device *sb_dev); 2153 struct netdev_queue *netdev_core_pick_tx(struct net_device *dev, 2154 struct sk_buff *skb, 2155 struct net_device *sb_dev); 2156 2157 /* returns the headroom that the master device needs to take in account 2158 * when forwarding to this dev 2159 */ 2160 static inline unsigned netdev_get_fwd_headroom(struct net_device *dev) 2161 { 2162 return dev->priv_flags & IFF_PHONY_HEADROOM ? 0 : dev->needed_headroom; 2163 } 2164 2165 static inline void netdev_set_rx_headroom(struct net_device *dev, int new_hr) 2166 { 2167 if (dev->netdev_ops->ndo_set_rx_headroom) 2168 dev->netdev_ops->ndo_set_rx_headroom(dev, new_hr); 2169 } 2170 2171 /* set the device rx headroom to the dev's default */ 2172 static inline void netdev_reset_rx_headroom(struct net_device *dev) 2173 { 2174 netdev_set_rx_headroom(dev, -1); 2175 } 2176 2177 /* 2178 * Net namespace inlines 2179 */ 2180 static inline 2181 struct net *dev_net(const struct net_device *dev) 2182 { 2183 return read_pnet(&dev->nd_net); 2184 } 2185 2186 static inline 2187 void dev_net_set(struct net_device *dev, struct net *net) 2188 { 2189 write_pnet(&dev->nd_net, net); 2190 } 2191 2192 /** 2193 * netdev_priv - access network device private data 2194 * @dev: network device 2195 * 2196 * Get network device private data 2197 */ 2198 static inline void *netdev_priv(const struct net_device *dev) 2199 { 2200 return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN); 2201 } 2202 2203 /* Set the sysfs physical device reference for the network logical device 2204 * if set prior to registration will cause a symlink during initialization. 2205 */ 2206 #define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev)) 2207 2208 /* Set the sysfs device type for the network logical device to allow 2209 * fine-grained identification of different network device types. For 2210 * example Ethernet, Wireless LAN, Bluetooth, WiMAX etc. 2211 */ 2212 #define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype)) 2213 2214 /* Default NAPI poll() weight 2215 * Device drivers are strongly advised to not use bigger value 2216 */ 2217 #define NAPI_POLL_WEIGHT 64 2218 2219 /** 2220 * netif_napi_add - initialize a NAPI context 2221 * @dev: network device 2222 * @napi: NAPI context 2223 * @poll: polling function 2224 * @weight: default weight 2225 * 2226 * netif_napi_add() must be used to initialize a NAPI context prior to calling 2227 * *any* of the other NAPI-related functions. 2228 */ 2229 void netif_napi_add(struct net_device *dev, struct napi_struct *napi, 2230 int (*poll)(struct napi_struct *, int), int weight); 2231 2232 /** 2233 * netif_tx_napi_add - initialize a NAPI context 2234 * @dev: network device 2235 * @napi: NAPI context 2236 * @poll: polling function 2237 * @weight: default weight 2238 * 2239 * This variant of netif_napi_add() should be used from drivers using NAPI 2240 * to exclusively poll a TX queue. 2241 * This will avoid we add it into napi_hash[], thus polluting this hash table. 2242 */ 2243 static inline void netif_tx_napi_add(struct net_device *dev, 2244 struct napi_struct *napi, 2245 int (*poll)(struct napi_struct *, int), 2246 int weight) 2247 { 2248 set_bit(NAPI_STATE_NO_BUSY_POLL, &napi->state); 2249 netif_napi_add(dev, napi, poll, weight); 2250 } 2251 2252 /** 2253 * netif_napi_del - remove a NAPI context 2254 * @napi: NAPI context 2255 * 2256 * netif_napi_del() removes a NAPI context from the network device NAPI list 2257 */ 2258 void netif_napi_del(struct napi_struct *napi); 2259 2260 struct napi_gro_cb { 2261 /* Virtual address of skb_shinfo(skb)->frags[0].page + offset. */ 2262 void *frag0; 2263 2264 /* Length of frag0. */ 2265 unsigned int frag0_len; 2266 2267 /* This indicates where we are processing relative to skb->data. */ 2268 int data_offset; 2269 2270 /* This is non-zero if the packet cannot be merged with the new skb. */ 2271 u16 flush; 2272 2273 /* Save the IP ID here and check when we get to the transport layer */ 2274 u16 flush_id; 2275 2276 /* Number of segments aggregated. */ 2277 u16 count; 2278 2279 /* Start offset for remote checksum offload */ 2280 u16 gro_remcsum_start; 2281 2282 /* jiffies when first packet was created/queued */ 2283 unsigned long age; 2284 2285 /* Used in ipv6_gro_receive() and foo-over-udp */ 2286 u16 proto; 2287 2288 /* This is non-zero if the packet may be of the same flow. */ 2289 u8 same_flow:1; 2290 2291 /* Used in tunnel GRO receive */ 2292 u8 encap_mark:1; 2293 2294 /* GRO checksum is valid */ 2295 u8 csum_valid:1; 2296 2297 /* Number of checksums via CHECKSUM_UNNECESSARY */ 2298 u8 csum_cnt:3; 2299 2300 /* Free the skb? */ 2301 u8 free:2; 2302 #define NAPI_GRO_FREE 1 2303 #define NAPI_GRO_FREE_STOLEN_HEAD 2 2304 2305 /* Used in foo-over-udp, set in udp[46]_gro_receive */ 2306 u8 is_ipv6:1; 2307 2308 /* Used in GRE, set in fou/gue_gro_receive */ 2309 u8 is_fou:1; 2310 2311 /* Used to determine if flush_id can be ignored */ 2312 u8 is_atomic:1; 2313 2314 /* Number of gro_receive callbacks this packet already went through */ 2315 u8 recursion_counter:4; 2316 2317 /* 1 bit hole */ 2318 2319 /* used to support CHECKSUM_COMPLETE for tunneling protocols */ 2320 __wsum csum; 2321 2322 /* used in skb_gro_receive() slow path */ 2323 struct sk_buff *last; 2324 }; 2325 2326 #define NAPI_GRO_CB(skb) ((struct napi_gro_cb *)(skb)->cb) 2327 2328 #define GRO_RECURSION_LIMIT 15 2329 static inline int gro_recursion_inc_test(struct sk_buff *skb) 2330 { 2331 return ++NAPI_GRO_CB(skb)->recursion_counter == GRO_RECURSION_LIMIT; 2332 } 2333 2334 typedef struct sk_buff *(*gro_receive_t)(struct list_head *, struct sk_buff *); 2335 static inline struct sk_buff *call_gro_receive(gro_receive_t cb, 2336 struct list_head *head, 2337 struct sk_buff *skb) 2338 { 2339 if (unlikely(gro_recursion_inc_test(skb))) { 2340 NAPI_GRO_CB(skb)->flush |= 1; 2341 return NULL; 2342 } 2343 2344 return cb(head, skb); 2345 } 2346 2347 typedef struct sk_buff *(*gro_receive_sk_t)(struct sock *, struct list_head *, 2348 struct sk_buff *); 2349 static inline struct sk_buff *call_gro_receive_sk(gro_receive_sk_t cb, 2350 struct sock *sk, 2351 struct list_head *head, 2352 struct sk_buff *skb) 2353 { 2354 if (unlikely(gro_recursion_inc_test(skb))) { 2355 NAPI_GRO_CB(skb)->flush |= 1; 2356 return NULL; 2357 } 2358 2359 return cb(sk, head, skb); 2360 } 2361 2362 struct packet_type { 2363 __be16 type; /* This is really htons(ether_type). */ 2364 bool ignore_outgoing; 2365 struct net_device *dev; /* NULL is wildcarded here */ 2366 int (*func) (struct sk_buff *, 2367 struct net_device *, 2368 struct packet_type *, 2369 struct net_device *); 2370 void (*list_func) (struct list_head *, 2371 struct packet_type *, 2372 struct net_device *); 2373 bool (*id_match)(struct packet_type *ptype, 2374 struct sock *sk); 2375 void *af_packet_priv; 2376 struct list_head list; 2377 }; 2378 2379 struct offload_callbacks { 2380 struct sk_buff *(*gso_segment)(struct sk_buff *skb, 2381 netdev_features_t features); 2382 struct sk_buff *(*gro_receive)(struct list_head *head, 2383 struct sk_buff *skb); 2384 int (*gro_complete)(struct sk_buff *skb, int nhoff); 2385 }; 2386 2387 struct packet_offload { 2388 __be16 type; /* This is really htons(ether_type). */ 2389 u16 priority; 2390 struct offload_callbacks callbacks; 2391 struct list_head list; 2392 }; 2393 2394 /* often modified stats are per-CPU, other are shared (netdev->stats) */ 2395 struct pcpu_sw_netstats { 2396 u64 rx_packets; 2397 u64 rx_bytes; 2398 u64 tx_packets; 2399 u64 tx_bytes; 2400 struct u64_stats_sync syncp; 2401 } __aligned(4 * sizeof(u64)); 2402 2403 struct pcpu_lstats { 2404 u64_stats_t packets; 2405 u64_stats_t bytes; 2406 struct u64_stats_sync syncp; 2407 } __aligned(2 * sizeof(u64)); 2408 2409 void dev_lstats_read(struct net_device *dev, u64 *packets, u64 *bytes); 2410 2411 static inline void dev_lstats_add(struct net_device *dev, unsigned int len) 2412 { 2413 struct pcpu_lstats *lstats = this_cpu_ptr(dev->lstats); 2414 2415 u64_stats_update_begin(&lstats->syncp); 2416 u64_stats_add(&lstats->bytes, len); 2417 u64_stats_inc(&lstats->packets); 2418 u64_stats_update_end(&lstats->syncp); 2419 } 2420 2421 #define __netdev_alloc_pcpu_stats(type, gfp) \ 2422 ({ \ 2423 typeof(type) __percpu *pcpu_stats = alloc_percpu_gfp(type, gfp);\ 2424 if (pcpu_stats) { \ 2425 int __cpu; \ 2426 for_each_possible_cpu(__cpu) { \ 2427 typeof(type) *stat; \ 2428 stat = per_cpu_ptr(pcpu_stats, __cpu); \ 2429 u64_stats_init(&stat->syncp); \ 2430 } \ 2431 } \ 2432 pcpu_stats; \ 2433 }) 2434 2435 #define netdev_alloc_pcpu_stats(type) \ 2436 __netdev_alloc_pcpu_stats(type, GFP_KERNEL) 2437 2438 enum netdev_lag_tx_type { 2439 NETDEV_LAG_TX_TYPE_UNKNOWN, 2440 NETDEV_LAG_TX_TYPE_RANDOM, 2441 NETDEV_LAG_TX_TYPE_BROADCAST, 2442 NETDEV_LAG_TX_TYPE_ROUNDROBIN, 2443 NETDEV_LAG_TX_TYPE_ACTIVEBACKUP, 2444 NETDEV_LAG_TX_TYPE_HASH, 2445 }; 2446 2447 enum netdev_lag_hash { 2448 NETDEV_LAG_HASH_NONE, 2449 NETDEV_LAG_HASH_L2, 2450 NETDEV_LAG_HASH_L34, 2451 NETDEV_LAG_HASH_L23, 2452 NETDEV_LAG_HASH_E23, 2453 NETDEV_LAG_HASH_E34, 2454 NETDEV_LAG_HASH_UNKNOWN, 2455 }; 2456 2457 struct netdev_lag_upper_info { 2458 enum netdev_lag_tx_type tx_type; 2459 enum netdev_lag_hash hash_type; 2460 }; 2461 2462 struct netdev_lag_lower_state_info { 2463 u8 link_up : 1, 2464 tx_enabled : 1; 2465 }; 2466 2467 #include <linux/notifier.h> 2468 2469 /* netdevice notifier chain. Please remember to update netdev_cmd_to_name() 2470 * and the rtnetlink notification exclusion list in rtnetlink_event() when 2471 * adding new types. 2472 */ 2473 enum netdev_cmd { 2474 NETDEV_UP = 1, /* For now you can't veto a device up/down */ 2475 NETDEV_DOWN, 2476 NETDEV_REBOOT, /* Tell a protocol stack a network interface 2477 detected a hardware crash and restarted 2478 - we can use this eg to kick tcp sessions 2479 once done */ 2480 NETDEV_CHANGE, /* Notify device state change */ 2481 NETDEV_REGISTER, 2482 NETDEV_UNREGISTER, 2483 NETDEV_CHANGEMTU, /* notify after mtu change happened */ 2484 NETDEV_CHANGEADDR, /* notify after the address change */ 2485 NETDEV_PRE_CHANGEADDR, /* notify before the address change */ 2486 NETDEV_GOING_DOWN, 2487 NETDEV_CHANGENAME, 2488 NETDEV_FEAT_CHANGE, 2489 NETDEV_BONDING_FAILOVER, 2490 NETDEV_PRE_UP, 2491 NETDEV_PRE_TYPE_CHANGE, 2492 NETDEV_POST_TYPE_CHANGE, 2493 NETDEV_POST_INIT, 2494 NETDEV_RELEASE, 2495 NETDEV_NOTIFY_PEERS, 2496 NETDEV_JOIN, 2497 NETDEV_CHANGEUPPER, 2498 NETDEV_RESEND_IGMP, 2499 NETDEV_PRECHANGEMTU, /* notify before mtu change happened */ 2500 NETDEV_CHANGEINFODATA, 2501 NETDEV_BONDING_INFO, 2502 NETDEV_PRECHANGEUPPER, 2503 NETDEV_CHANGELOWERSTATE, 2504 NETDEV_UDP_TUNNEL_PUSH_INFO, 2505 NETDEV_UDP_TUNNEL_DROP_INFO, 2506 NETDEV_CHANGE_TX_QUEUE_LEN, 2507 NETDEV_CVLAN_FILTER_PUSH_INFO, 2508 NETDEV_CVLAN_FILTER_DROP_INFO, 2509 NETDEV_SVLAN_FILTER_PUSH_INFO, 2510 NETDEV_SVLAN_FILTER_DROP_INFO, 2511 }; 2512 const char *netdev_cmd_to_name(enum netdev_cmd cmd); 2513 2514 int register_netdevice_notifier(struct notifier_block *nb); 2515 int unregister_netdevice_notifier(struct notifier_block *nb); 2516 int register_netdevice_notifier_net(struct net *net, struct notifier_block *nb); 2517 int unregister_netdevice_notifier_net(struct net *net, 2518 struct notifier_block *nb); 2519 2520 struct netdev_notifier_info { 2521 struct net_device *dev; 2522 struct netlink_ext_ack *extack; 2523 }; 2524 2525 struct netdev_notifier_info_ext { 2526 struct netdev_notifier_info info; /* must be first */ 2527 union { 2528 u32 mtu; 2529 } ext; 2530 }; 2531 2532 struct netdev_notifier_change_info { 2533 struct netdev_notifier_info info; /* must be first */ 2534 unsigned int flags_changed; 2535 }; 2536 2537 struct netdev_notifier_changeupper_info { 2538 struct netdev_notifier_info info; /* must be first */ 2539 struct net_device *upper_dev; /* new upper dev */ 2540 bool master; /* is upper dev master */ 2541 bool linking; /* is the notification for link or unlink */ 2542 void *upper_info; /* upper dev info */ 2543 }; 2544 2545 struct netdev_notifier_changelowerstate_info { 2546 struct netdev_notifier_info info; /* must be first */ 2547 void *lower_state_info; /* is lower dev state */ 2548 }; 2549 2550 struct netdev_notifier_pre_changeaddr_info { 2551 struct netdev_notifier_info info; /* must be first */ 2552 const unsigned char *dev_addr; 2553 }; 2554 2555 static inline void netdev_notifier_info_init(struct netdev_notifier_info *info, 2556 struct net_device *dev) 2557 { 2558 info->dev = dev; 2559 info->extack = NULL; 2560 } 2561 2562 static inline struct net_device * 2563 netdev_notifier_info_to_dev(const struct netdev_notifier_info *info) 2564 { 2565 return info->dev; 2566 } 2567 2568 static inline struct netlink_ext_ack * 2569 netdev_notifier_info_to_extack(const struct netdev_notifier_info *info) 2570 { 2571 return info->extack; 2572 } 2573 2574 int call_netdevice_notifiers(unsigned long val, struct net_device *dev); 2575 2576 2577 extern rwlock_t dev_base_lock; /* Device list lock */ 2578 2579 #define for_each_netdev(net, d) \ 2580 list_for_each_entry(d, &(net)->dev_base_head, dev_list) 2581 #define for_each_netdev_reverse(net, d) \ 2582 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list) 2583 #define for_each_netdev_rcu(net, d) \ 2584 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list) 2585 #define for_each_netdev_safe(net, d, n) \ 2586 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list) 2587 #define for_each_netdev_continue(net, d) \ 2588 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list) 2589 #define for_each_netdev_continue_reverse(net, d) \ 2590 list_for_each_entry_continue_reverse(d, &(net)->dev_base_head, \ 2591 dev_list) 2592 #define for_each_netdev_continue_rcu(net, d) \ 2593 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list) 2594 #define for_each_netdev_in_bond_rcu(bond, slave) \ 2595 for_each_netdev_rcu(&init_net, slave) \ 2596 if (netdev_master_upper_dev_get_rcu(slave) == (bond)) 2597 #define net_device_entry(lh) list_entry(lh, struct net_device, dev_list) 2598 2599 static inline struct net_device *next_net_device(struct net_device *dev) 2600 { 2601 struct list_head *lh; 2602 struct net *net; 2603 2604 net = dev_net(dev); 2605 lh = dev->dev_list.next; 2606 return lh == &net->dev_base_head ? NULL : net_device_entry(lh); 2607 } 2608 2609 static inline struct net_device *next_net_device_rcu(struct net_device *dev) 2610 { 2611 struct list_head *lh; 2612 struct net *net; 2613 2614 net = dev_net(dev); 2615 lh = rcu_dereference(list_next_rcu(&dev->dev_list)); 2616 return lh == &net->dev_base_head ? NULL : net_device_entry(lh); 2617 } 2618 2619 static inline struct net_device *first_net_device(struct net *net) 2620 { 2621 return list_empty(&net->dev_base_head) ? NULL : 2622 net_device_entry(net->dev_base_head.next); 2623 } 2624 2625 static inline struct net_device *first_net_device_rcu(struct net *net) 2626 { 2627 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head)); 2628 2629 return lh == &net->dev_base_head ? NULL : net_device_entry(lh); 2630 } 2631 2632 int netdev_boot_setup_check(struct net_device *dev); 2633 unsigned long netdev_boot_base(const char *prefix, int unit); 2634 struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type, 2635 const char *hwaddr); 2636 struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type); 2637 struct net_device *__dev_getfirstbyhwtype(struct net *net, unsigned short type); 2638 void dev_add_pack(struct packet_type *pt); 2639 void dev_remove_pack(struct packet_type *pt); 2640 void __dev_remove_pack(struct packet_type *pt); 2641 void dev_add_offload(struct packet_offload *po); 2642 void dev_remove_offload(struct packet_offload *po); 2643 2644 int dev_get_iflink(const struct net_device *dev); 2645 int dev_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb); 2646 struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags, 2647 unsigned short mask); 2648 struct net_device *dev_get_by_name(struct net *net, const char *name); 2649 struct net_device *dev_get_by_name_rcu(struct net *net, const char *name); 2650 struct net_device *__dev_get_by_name(struct net *net, const char *name); 2651 int dev_alloc_name(struct net_device *dev, const char *name); 2652 int dev_open(struct net_device *dev, struct netlink_ext_ack *extack); 2653 void dev_close(struct net_device *dev); 2654 void dev_close_many(struct list_head *head, bool unlink); 2655 void dev_disable_lro(struct net_device *dev); 2656 int dev_loopback_xmit(struct net *net, struct sock *sk, struct sk_buff *newskb); 2657 u16 dev_pick_tx_zero(struct net_device *dev, struct sk_buff *skb, 2658 struct net_device *sb_dev); 2659 u16 dev_pick_tx_cpu_id(struct net_device *dev, struct sk_buff *skb, 2660 struct net_device *sb_dev); 2661 int dev_queue_xmit(struct sk_buff *skb); 2662 int dev_queue_xmit_accel(struct sk_buff *skb, struct net_device *sb_dev); 2663 int dev_direct_xmit(struct sk_buff *skb, u16 queue_id); 2664 int register_netdevice(struct net_device *dev); 2665 void unregister_netdevice_queue(struct net_device *dev, struct list_head *head); 2666 void unregister_netdevice_many(struct list_head *head); 2667 static inline void unregister_netdevice(struct net_device *dev) 2668 { 2669 unregister_netdevice_queue(dev, NULL); 2670 } 2671 2672 int netdev_refcnt_read(const struct net_device *dev); 2673 void free_netdev(struct net_device *dev); 2674 void netdev_freemem(struct net_device *dev); 2675 void synchronize_net(void); 2676 int init_dummy_netdev(struct net_device *dev); 2677 2678 struct net_device *dev_get_by_index(struct net *net, int ifindex); 2679 struct net_device *__dev_get_by_index(struct net *net, int ifindex); 2680 struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex); 2681 struct net_device *dev_get_by_napi_id(unsigned int napi_id); 2682 int netdev_get_name(struct net *net, char *name, int ifindex); 2683 int dev_restart(struct net_device *dev); 2684 int skb_gro_receive(struct sk_buff *p, struct sk_buff *skb); 2685 2686 static inline unsigned int skb_gro_offset(const struct sk_buff *skb) 2687 { 2688 return NAPI_GRO_CB(skb)->data_offset; 2689 } 2690 2691 static inline unsigned int skb_gro_len(const struct sk_buff *skb) 2692 { 2693 return skb->len - NAPI_GRO_CB(skb)->data_offset; 2694 } 2695 2696 static inline void skb_gro_pull(struct sk_buff *skb, unsigned int len) 2697 { 2698 NAPI_GRO_CB(skb)->data_offset += len; 2699 } 2700 2701 static inline void *skb_gro_header_fast(struct sk_buff *skb, 2702 unsigned int offset) 2703 { 2704 return NAPI_GRO_CB(skb)->frag0 + offset; 2705 } 2706 2707 static inline int skb_gro_header_hard(struct sk_buff *skb, unsigned int hlen) 2708 { 2709 return NAPI_GRO_CB(skb)->frag0_len < hlen; 2710 } 2711 2712 static inline void skb_gro_frag0_invalidate(struct sk_buff *skb) 2713 { 2714 NAPI_GRO_CB(skb)->frag0 = NULL; 2715 NAPI_GRO_CB(skb)->frag0_len = 0; 2716 } 2717 2718 static inline void *skb_gro_header_slow(struct sk_buff *skb, unsigned int hlen, 2719 unsigned int offset) 2720 { 2721 if (!pskb_may_pull(skb, hlen)) 2722 return NULL; 2723 2724 skb_gro_frag0_invalidate(skb); 2725 return skb->data + offset; 2726 } 2727 2728 static inline void *skb_gro_network_header(struct sk_buff *skb) 2729 { 2730 return (NAPI_GRO_CB(skb)->frag0 ?: skb->data) + 2731 skb_network_offset(skb); 2732 } 2733 2734 static inline void skb_gro_postpull_rcsum(struct sk_buff *skb, 2735 const void *start, unsigned int len) 2736 { 2737 if (NAPI_GRO_CB(skb)->csum_valid) 2738 NAPI_GRO_CB(skb)->csum = csum_sub(NAPI_GRO_CB(skb)->csum, 2739 csum_partial(start, len, 0)); 2740 } 2741 2742 /* GRO checksum functions. These are logical equivalents of the normal 2743 * checksum functions (in skbuff.h) except that they operate on the GRO 2744 * offsets and fields in sk_buff. 2745 */ 2746 2747 __sum16 __skb_gro_checksum_complete(struct sk_buff *skb); 2748 2749 static inline bool skb_at_gro_remcsum_start(struct sk_buff *skb) 2750 { 2751 return (NAPI_GRO_CB(skb)->gro_remcsum_start == skb_gro_offset(skb)); 2752 } 2753 2754 static inline bool __skb_gro_checksum_validate_needed(struct sk_buff *skb, 2755 bool zero_okay, 2756 __sum16 check) 2757 { 2758 return ((skb->ip_summed != CHECKSUM_PARTIAL || 2759 skb_checksum_start_offset(skb) < 2760 skb_gro_offset(skb)) && 2761 !skb_at_gro_remcsum_start(skb) && 2762 NAPI_GRO_CB(skb)->csum_cnt == 0 && 2763 (!zero_okay || check)); 2764 } 2765 2766 static inline __sum16 __skb_gro_checksum_validate_complete(struct sk_buff *skb, 2767 __wsum psum) 2768 { 2769 if (NAPI_GRO_CB(skb)->csum_valid && 2770 !csum_fold(csum_add(psum, NAPI_GRO_CB(skb)->csum))) 2771 return 0; 2772 2773 NAPI_GRO_CB(skb)->csum = psum; 2774 2775 return __skb_gro_checksum_complete(skb); 2776 } 2777 2778 static inline void skb_gro_incr_csum_unnecessary(struct sk_buff *skb) 2779 { 2780 if (NAPI_GRO_CB(skb)->csum_cnt > 0) { 2781 /* Consume a checksum from CHECKSUM_UNNECESSARY */ 2782 NAPI_GRO_CB(skb)->csum_cnt--; 2783 } else { 2784 /* Update skb for CHECKSUM_UNNECESSARY and csum_level when we 2785 * verified a new top level checksum or an encapsulated one 2786 * during GRO. This saves work if we fallback to normal path. 2787 */ 2788 __skb_incr_checksum_unnecessary(skb); 2789 } 2790 } 2791 2792 #define __skb_gro_checksum_validate(skb, proto, zero_okay, check, \ 2793 compute_pseudo) \ 2794 ({ \ 2795 __sum16 __ret = 0; \ 2796 if (__skb_gro_checksum_validate_needed(skb, zero_okay, check)) \ 2797 __ret = __skb_gro_checksum_validate_complete(skb, \ 2798 compute_pseudo(skb, proto)); \ 2799 if (!__ret) \ 2800 skb_gro_incr_csum_unnecessary(skb); \ 2801 __ret; \ 2802 }) 2803 2804 #define skb_gro_checksum_validate(skb, proto, compute_pseudo) \ 2805 __skb_gro_checksum_validate(skb, proto, false, 0, compute_pseudo) 2806 2807 #define skb_gro_checksum_validate_zero_check(skb, proto, check, \ 2808 compute_pseudo) \ 2809 __skb_gro_checksum_validate(skb, proto, true, check, compute_pseudo) 2810 2811 #define skb_gro_checksum_simple_validate(skb) \ 2812 __skb_gro_checksum_validate(skb, 0, false, 0, null_compute_pseudo) 2813 2814 static inline bool __skb_gro_checksum_convert_check(struct sk_buff *skb) 2815 { 2816 return (NAPI_GRO_CB(skb)->csum_cnt == 0 && 2817 !NAPI_GRO_CB(skb)->csum_valid); 2818 } 2819 2820 static inline void __skb_gro_checksum_convert(struct sk_buff *skb, 2821 __sum16 check, __wsum pseudo) 2822 { 2823 NAPI_GRO_CB(skb)->csum = ~pseudo; 2824 NAPI_GRO_CB(skb)->csum_valid = 1; 2825 } 2826 2827 #define skb_gro_checksum_try_convert(skb, proto, check, compute_pseudo) \ 2828 do { \ 2829 if (__skb_gro_checksum_convert_check(skb)) \ 2830 __skb_gro_checksum_convert(skb, check, \ 2831 compute_pseudo(skb, proto)); \ 2832 } while (0) 2833 2834 struct gro_remcsum { 2835 int offset; 2836 __wsum delta; 2837 }; 2838 2839 static inline void skb_gro_remcsum_init(struct gro_remcsum *grc) 2840 { 2841 grc->offset = 0; 2842 grc->delta = 0; 2843 } 2844 2845 static inline void *skb_gro_remcsum_process(struct sk_buff *skb, void *ptr, 2846 unsigned int off, size_t hdrlen, 2847 int start, int offset, 2848 struct gro_remcsum *grc, 2849 bool nopartial) 2850 { 2851 __wsum delta; 2852 size_t plen = hdrlen + max_t(size_t, offset + sizeof(u16), start); 2853 2854 BUG_ON(!NAPI_GRO_CB(skb)->csum_valid); 2855 2856 if (!nopartial) { 2857 NAPI_GRO_CB(skb)->gro_remcsum_start = off + hdrlen + start; 2858 return ptr; 2859 } 2860 2861 ptr = skb_gro_header_fast(skb, off); 2862 if (skb_gro_header_hard(skb, off + plen)) { 2863 ptr = skb_gro_header_slow(skb, off + plen, off); 2864 if (!ptr) 2865 return NULL; 2866 } 2867 2868 delta = remcsum_adjust(ptr + hdrlen, NAPI_GRO_CB(skb)->csum, 2869 start, offset); 2870 2871 /* Adjust skb->csum since we changed the packet */ 2872 NAPI_GRO_CB(skb)->csum = csum_add(NAPI_GRO_CB(skb)->csum, delta); 2873 2874 grc->offset = off + hdrlen + offset; 2875 grc->delta = delta; 2876 2877 return ptr; 2878 } 2879 2880 static inline void skb_gro_remcsum_cleanup(struct sk_buff *skb, 2881 struct gro_remcsum *grc) 2882 { 2883 void *ptr; 2884 size_t plen = grc->offset + sizeof(u16); 2885 2886 if (!grc->delta) 2887 return; 2888 2889 ptr = skb_gro_header_fast(skb, grc->offset); 2890 if (skb_gro_header_hard(skb, grc->offset + sizeof(u16))) { 2891 ptr = skb_gro_header_slow(skb, plen, grc->offset); 2892 if (!ptr) 2893 return; 2894 } 2895 2896 remcsum_unadjust((__sum16 *)ptr, grc->delta); 2897 } 2898 2899 #ifdef CONFIG_XFRM_OFFLOAD 2900 static inline void skb_gro_flush_final(struct sk_buff *skb, struct sk_buff *pp, int flush) 2901 { 2902 if (PTR_ERR(pp) != -EINPROGRESS) 2903 NAPI_GRO_CB(skb)->flush |= flush; 2904 } 2905 static inline void skb_gro_flush_final_remcsum(struct sk_buff *skb, 2906 struct sk_buff *pp, 2907 int flush, 2908 struct gro_remcsum *grc) 2909 { 2910 if (PTR_ERR(pp) != -EINPROGRESS) { 2911 NAPI_GRO_CB(skb)->flush |= flush; 2912 skb_gro_remcsum_cleanup(skb, grc); 2913 skb->remcsum_offload = 0; 2914 } 2915 } 2916 #else 2917 static inline void skb_gro_flush_final(struct sk_buff *skb, struct sk_buff *pp, int flush) 2918 { 2919 NAPI_GRO_CB(skb)->flush |= flush; 2920 } 2921 static inline void skb_gro_flush_final_remcsum(struct sk_buff *skb, 2922 struct sk_buff *pp, 2923 int flush, 2924 struct gro_remcsum *grc) 2925 { 2926 NAPI_GRO_CB(skb)->flush |= flush; 2927 skb_gro_remcsum_cleanup(skb, grc); 2928 skb->remcsum_offload = 0; 2929 } 2930 #endif 2931 2932 static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev, 2933 unsigned short type, 2934 const void *daddr, const void *saddr, 2935 unsigned int len) 2936 { 2937 if (!dev->header_ops || !dev->header_ops->create) 2938 return 0; 2939 2940 return dev->header_ops->create(skb, dev, type, daddr, saddr, len); 2941 } 2942 2943 static inline int dev_parse_header(const struct sk_buff *skb, 2944 unsigned char *haddr) 2945 { 2946 const struct net_device *dev = skb->dev; 2947 2948 if (!dev->header_ops || !dev->header_ops->parse) 2949 return 0; 2950 return dev->header_ops->parse(skb, haddr); 2951 } 2952 2953 static inline __be16 dev_parse_header_protocol(const struct sk_buff *skb) 2954 { 2955 const struct net_device *dev = skb->dev; 2956 2957 if (!dev->header_ops || !dev->header_ops->parse_protocol) 2958 return 0; 2959 return dev->header_ops->parse_protocol(skb); 2960 } 2961 2962 /* ll_header must have at least hard_header_len allocated */ 2963 static inline bool dev_validate_header(const struct net_device *dev, 2964 char *ll_header, int len) 2965 { 2966 if (likely(len >= dev->hard_header_len)) 2967 return true; 2968 if (len < dev->min_header_len) 2969 return false; 2970 2971 if (capable(CAP_SYS_RAWIO)) { 2972 memset(ll_header + len, 0, dev->hard_header_len - len); 2973 return true; 2974 } 2975 2976 if (dev->header_ops && dev->header_ops->validate) 2977 return dev->header_ops->validate(ll_header, len); 2978 2979 return false; 2980 } 2981 2982 typedef int gifconf_func_t(struct net_device * dev, char __user * bufptr, 2983 int len, int size); 2984 int register_gifconf(unsigned int family, gifconf_func_t *gifconf); 2985 static inline int unregister_gifconf(unsigned int family) 2986 { 2987 return register_gifconf(family, NULL); 2988 } 2989 2990 #ifdef CONFIG_NET_FLOW_LIMIT 2991 #define FLOW_LIMIT_HISTORY (1 << 7) /* must be ^2 and !overflow buckets */ 2992 struct sd_flow_limit { 2993 u64 count; 2994 unsigned int num_buckets; 2995 unsigned int history_head; 2996 u16 history[FLOW_LIMIT_HISTORY]; 2997 u8 buckets[]; 2998 }; 2999 3000 extern int netdev_flow_limit_table_len; 3001 #endif /* CONFIG_NET_FLOW_LIMIT */ 3002 3003 /* 3004 * Incoming packets are placed on per-CPU queues 3005 */ 3006 struct softnet_data { 3007 struct list_head poll_list; 3008 struct sk_buff_head process_queue; 3009 3010 /* stats */ 3011 unsigned int processed; 3012 unsigned int time_squeeze; 3013 unsigned int received_rps; 3014 #ifdef CONFIG_RPS 3015 struct softnet_data *rps_ipi_list; 3016 #endif 3017 #ifdef CONFIG_NET_FLOW_LIMIT 3018 struct sd_flow_limit __rcu *flow_limit; 3019 #endif 3020 struct Qdisc *output_queue; 3021 struct Qdisc **output_queue_tailp; 3022 struct sk_buff *completion_queue; 3023 #ifdef CONFIG_XFRM_OFFLOAD 3024 struct sk_buff_head xfrm_backlog; 3025 #endif 3026 /* written and read only by owning cpu: */ 3027 struct { 3028 u16 recursion; 3029 u8 more; 3030 } xmit; 3031 #ifdef CONFIG_RPS 3032 /* input_queue_head should be written by cpu owning this struct, 3033 * and only read by other cpus. Worth using a cache line. 3034 */ 3035 unsigned int input_queue_head ____cacheline_aligned_in_smp; 3036 3037 /* Elements below can be accessed between CPUs for RPS/RFS */ 3038 call_single_data_t csd ____cacheline_aligned_in_smp; 3039 struct softnet_data *rps_ipi_next; 3040 unsigned int cpu; 3041 unsigned int input_queue_tail; 3042 #endif 3043 unsigned int dropped; 3044 struct sk_buff_head input_pkt_queue; 3045 struct napi_struct backlog; 3046 3047 }; 3048 3049 static inline void input_queue_head_incr(struct softnet_data *sd) 3050 { 3051 #ifdef CONFIG_RPS 3052 sd->input_queue_head++; 3053 #endif 3054 } 3055 3056 static inline void input_queue_tail_incr_save(struct softnet_data *sd, 3057 unsigned int *qtail) 3058 { 3059 #ifdef CONFIG_RPS 3060 *qtail = ++sd->input_queue_tail; 3061 #endif 3062 } 3063 3064 DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data); 3065 3066 static inline int dev_recursion_level(void) 3067 { 3068 return this_cpu_read(softnet_data.xmit.recursion); 3069 } 3070 3071 #define XMIT_RECURSION_LIMIT 10 3072 static inline bool dev_xmit_recursion(void) 3073 { 3074 return unlikely(__this_cpu_read(softnet_data.xmit.recursion) > 3075 XMIT_RECURSION_LIMIT); 3076 } 3077 3078 static inline void dev_xmit_recursion_inc(void) 3079 { 3080 __this_cpu_inc(softnet_data.xmit.recursion); 3081 } 3082 3083 static inline void dev_xmit_recursion_dec(void) 3084 { 3085 __this_cpu_dec(softnet_data.xmit.recursion); 3086 } 3087 3088 void __netif_schedule(struct Qdisc *q); 3089 void netif_schedule_queue(struct netdev_queue *txq); 3090 3091 static inline void netif_tx_schedule_all(struct net_device *dev) 3092 { 3093 unsigned int i; 3094 3095 for (i = 0; i < dev->num_tx_queues; i++) 3096 netif_schedule_queue(netdev_get_tx_queue(dev, i)); 3097 } 3098 3099 static __always_inline void netif_tx_start_queue(struct netdev_queue *dev_queue) 3100 { 3101 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state); 3102 } 3103 3104 /** 3105 * netif_start_queue - allow transmit 3106 * @dev: network device 3107 * 3108 * Allow upper layers to call the device hard_start_xmit routine. 3109 */ 3110 static inline void netif_start_queue(struct net_device *dev) 3111 { 3112 netif_tx_start_queue(netdev_get_tx_queue(dev, 0)); 3113 } 3114 3115 static inline void netif_tx_start_all_queues(struct net_device *dev) 3116 { 3117 unsigned int i; 3118 3119 for (i = 0; i < dev->num_tx_queues; i++) { 3120 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 3121 netif_tx_start_queue(txq); 3122 } 3123 } 3124 3125 void netif_tx_wake_queue(struct netdev_queue *dev_queue); 3126 3127 /** 3128 * netif_wake_queue - restart transmit 3129 * @dev: network device 3130 * 3131 * Allow upper layers to call the device hard_start_xmit routine. 3132 * Used for flow control when transmit resources are available. 3133 */ 3134 static inline void netif_wake_queue(struct net_device *dev) 3135 { 3136 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0)); 3137 } 3138 3139 static inline void netif_tx_wake_all_queues(struct net_device *dev) 3140 { 3141 unsigned int i; 3142 3143 for (i = 0; i < dev->num_tx_queues; i++) { 3144 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 3145 netif_tx_wake_queue(txq); 3146 } 3147 } 3148 3149 static __always_inline void netif_tx_stop_queue(struct netdev_queue *dev_queue) 3150 { 3151 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state); 3152 } 3153 3154 /** 3155 * netif_stop_queue - stop transmitted packets 3156 * @dev: network device 3157 * 3158 * Stop upper layers calling the device hard_start_xmit routine. 3159 * Used for flow control when transmit resources are unavailable. 3160 */ 3161 static inline void netif_stop_queue(struct net_device *dev) 3162 { 3163 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0)); 3164 } 3165 3166 void netif_tx_stop_all_queues(struct net_device *dev); 3167 void netdev_update_lockdep_key(struct net_device *dev); 3168 3169 static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue) 3170 { 3171 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state); 3172 } 3173 3174 /** 3175 * netif_queue_stopped - test if transmit queue is flowblocked 3176 * @dev: network device 3177 * 3178 * Test if transmit queue on device is currently unable to send. 3179 */ 3180 static inline bool netif_queue_stopped(const struct net_device *dev) 3181 { 3182 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0)); 3183 } 3184 3185 static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue) 3186 { 3187 return dev_queue->state & QUEUE_STATE_ANY_XOFF; 3188 } 3189 3190 static inline bool 3191 netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue) 3192 { 3193 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN; 3194 } 3195 3196 static inline bool 3197 netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue) 3198 { 3199 return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN; 3200 } 3201 3202 /** 3203 * netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write 3204 * @dev_queue: pointer to transmit queue 3205 * 3206 * BQL enabled drivers might use this helper in their ndo_start_xmit(), 3207 * to give appropriate hint to the CPU. 3208 */ 3209 static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue) 3210 { 3211 #ifdef CONFIG_BQL 3212 prefetchw(&dev_queue->dql.num_queued); 3213 #endif 3214 } 3215 3216 /** 3217 * netdev_txq_bql_complete_prefetchw - prefetch bql data for write 3218 * @dev_queue: pointer to transmit queue 3219 * 3220 * BQL enabled drivers might use this helper in their TX completion path, 3221 * to give appropriate hint to the CPU. 3222 */ 3223 static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue) 3224 { 3225 #ifdef CONFIG_BQL 3226 prefetchw(&dev_queue->dql.limit); 3227 #endif 3228 } 3229 3230 static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue, 3231 unsigned int bytes) 3232 { 3233 #ifdef CONFIG_BQL 3234 dql_queued(&dev_queue->dql, bytes); 3235 3236 if (likely(dql_avail(&dev_queue->dql) >= 0)) 3237 return; 3238 3239 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state); 3240 3241 /* 3242 * The XOFF flag must be set before checking the dql_avail below, 3243 * because in netdev_tx_completed_queue we update the dql_completed 3244 * before checking the XOFF flag. 3245 */ 3246 smp_mb(); 3247 3248 /* check again in case another CPU has just made room avail */ 3249 if (unlikely(dql_avail(&dev_queue->dql) >= 0)) 3250 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state); 3251 #endif 3252 } 3253 3254 /* Variant of netdev_tx_sent_queue() for drivers that are aware 3255 * that they should not test BQL status themselves. 3256 * We do want to change __QUEUE_STATE_STACK_XOFF only for the last 3257 * skb of a batch. 3258 * Returns true if the doorbell must be used to kick the NIC. 3259 */ 3260 static inline bool __netdev_tx_sent_queue(struct netdev_queue *dev_queue, 3261 unsigned int bytes, 3262 bool xmit_more) 3263 { 3264 if (xmit_more) { 3265 #ifdef CONFIG_BQL 3266 dql_queued(&dev_queue->dql, bytes); 3267 #endif 3268 return netif_tx_queue_stopped(dev_queue); 3269 } 3270 netdev_tx_sent_queue(dev_queue, bytes); 3271 return true; 3272 } 3273 3274 /** 3275 * netdev_sent_queue - report the number of bytes queued to hardware 3276 * @dev: network device 3277 * @bytes: number of bytes queued to the hardware device queue 3278 * 3279 * Report the number of bytes queued for sending/completion to the network 3280 * device hardware queue. @bytes should be a good approximation and should 3281 * exactly match netdev_completed_queue() @bytes 3282 */ 3283 static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes) 3284 { 3285 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes); 3286 } 3287 3288 static inline bool __netdev_sent_queue(struct net_device *dev, 3289 unsigned int bytes, 3290 bool xmit_more) 3291 { 3292 return __netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes, 3293 xmit_more); 3294 } 3295 3296 static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue, 3297 unsigned int pkts, unsigned int bytes) 3298 { 3299 #ifdef CONFIG_BQL 3300 if (unlikely(!bytes)) 3301 return; 3302 3303 dql_completed(&dev_queue->dql, bytes); 3304 3305 /* 3306 * Without the memory barrier there is a small possiblity that 3307 * netdev_tx_sent_queue will miss the update and cause the queue to 3308 * be stopped forever 3309 */ 3310 smp_mb(); 3311 3312 if (unlikely(dql_avail(&dev_queue->dql) < 0)) 3313 return; 3314 3315 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state)) 3316 netif_schedule_queue(dev_queue); 3317 #endif 3318 } 3319 3320 /** 3321 * netdev_completed_queue - report bytes and packets completed by device 3322 * @dev: network device 3323 * @pkts: actual number of packets sent over the medium 3324 * @bytes: actual number of bytes sent over the medium 3325 * 3326 * Report the number of bytes and packets transmitted by the network device 3327 * hardware queue over the physical medium, @bytes must exactly match the 3328 * @bytes amount passed to netdev_sent_queue() 3329 */ 3330 static inline void netdev_completed_queue(struct net_device *dev, 3331 unsigned int pkts, unsigned int bytes) 3332 { 3333 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes); 3334 } 3335 3336 static inline void netdev_tx_reset_queue(struct netdev_queue *q) 3337 { 3338 #ifdef CONFIG_BQL 3339 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state); 3340 dql_reset(&q->dql); 3341 #endif 3342 } 3343 3344 /** 3345 * netdev_reset_queue - reset the packets and bytes count of a network device 3346 * @dev_queue: network device 3347 * 3348 * Reset the bytes and packet count of a network device and clear the 3349 * software flow control OFF bit for this network device 3350 */ 3351 static inline void netdev_reset_queue(struct net_device *dev_queue) 3352 { 3353 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0)); 3354 } 3355 3356 /** 3357 * netdev_cap_txqueue - check if selected tx queue exceeds device queues 3358 * @dev: network device 3359 * @queue_index: given tx queue index 3360 * 3361 * Returns 0 if given tx queue index >= number of device tx queues, 3362 * otherwise returns the originally passed tx queue index. 3363 */ 3364 static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index) 3365 { 3366 if (unlikely(queue_index >= dev->real_num_tx_queues)) { 3367 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n", 3368 dev->name, queue_index, 3369 dev->real_num_tx_queues); 3370 return 0; 3371 } 3372 3373 return queue_index; 3374 } 3375 3376 /** 3377 * netif_running - test if up 3378 * @dev: network device 3379 * 3380 * Test if the device has been brought up. 3381 */ 3382 static inline bool netif_running(const struct net_device *dev) 3383 { 3384 return test_bit(__LINK_STATE_START, &dev->state); 3385 } 3386 3387 /* 3388 * Routines to manage the subqueues on a device. We only need start, 3389 * stop, and a check if it's stopped. All other device management is 3390 * done at the overall netdevice level. 3391 * Also test the device if we're multiqueue. 3392 */ 3393 3394 /** 3395 * netif_start_subqueue - allow sending packets on subqueue 3396 * @dev: network device 3397 * @queue_index: sub queue index 3398 * 3399 * Start individual transmit queue of a device with multiple transmit queues. 3400 */ 3401 static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index) 3402 { 3403 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 3404 3405 netif_tx_start_queue(txq); 3406 } 3407 3408 /** 3409 * netif_stop_subqueue - stop sending packets on subqueue 3410 * @dev: network device 3411 * @queue_index: sub queue index 3412 * 3413 * Stop individual transmit queue of a device with multiple transmit queues. 3414 */ 3415 static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index) 3416 { 3417 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 3418 netif_tx_stop_queue(txq); 3419 } 3420 3421 /** 3422 * netif_subqueue_stopped - test status of subqueue 3423 * @dev: network device 3424 * @queue_index: sub queue index 3425 * 3426 * Check individual transmit queue of a device with multiple transmit queues. 3427 */ 3428 static inline bool __netif_subqueue_stopped(const struct net_device *dev, 3429 u16 queue_index) 3430 { 3431 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 3432 3433 return netif_tx_queue_stopped(txq); 3434 } 3435 3436 static inline bool netif_subqueue_stopped(const struct net_device *dev, 3437 struct sk_buff *skb) 3438 { 3439 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb)); 3440 } 3441 3442 /** 3443 * netif_wake_subqueue - allow sending packets on subqueue 3444 * @dev: network device 3445 * @queue_index: sub queue index 3446 * 3447 * Resume individual transmit queue of a device with multiple transmit queues. 3448 */ 3449 static inline void netif_wake_subqueue(struct net_device *dev, u16 queue_index) 3450 { 3451 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 3452 3453 netif_tx_wake_queue(txq); 3454 } 3455 3456 #ifdef CONFIG_XPS 3457 int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask, 3458 u16 index); 3459 int __netif_set_xps_queue(struct net_device *dev, const unsigned long *mask, 3460 u16 index, bool is_rxqs_map); 3461 3462 /** 3463 * netif_attr_test_mask - Test a CPU or Rx queue set in a mask 3464 * @j: CPU/Rx queue index 3465 * @mask: bitmask of all cpus/rx queues 3466 * @nr_bits: number of bits in the bitmask 3467 * 3468 * Test if a CPU or Rx queue index is set in a mask of all CPU/Rx queues. 3469 */ 3470 static inline bool netif_attr_test_mask(unsigned long j, 3471 const unsigned long *mask, 3472 unsigned int nr_bits) 3473 { 3474 cpu_max_bits_warn(j, nr_bits); 3475 return test_bit(j, mask); 3476 } 3477 3478 /** 3479 * netif_attr_test_online - Test for online CPU/Rx queue 3480 * @j: CPU/Rx queue index 3481 * @online_mask: bitmask for CPUs/Rx queues that are online 3482 * @nr_bits: number of bits in the bitmask 3483 * 3484 * Returns true if a CPU/Rx queue is online. 3485 */ 3486 static inline bool netif_attr_test_online(unsigned long j, 3487 const unsigned long *online_mask, 3488 unsigned int nr_bits) 3489 { 3490 cpu_max_bits_warn(j, nr_bits); 3491 3492 if (online_mask) 3493 return test_bit(j, online_mask); 3494 3495 return (j < nr_bits); 3496 } 3497 3498 /** 3499 * netif_attrmask_next - get the next CPU/Rx queue in a cpu/Rx queues mask 3500 * @n: CPU/Rx queue index 3501 * @srcp: the cpumask/Rx queue mask pointer 3502 * @nr_bits: number of bits in the bitmask 3503 * 3504 * Returns >= nr_bits if no further CPUs/Rx queues set. 3505 */ 3506 static inline unsigned int netif_attrmask_next(int n, const unsigned long *srcp, 3507 unsigned int nr_bits) 3508 { 3509 /* -1 is a legal arg here. */ 3510 if (n != -1) 3511 cpu_max_bits_warn(n, nr_bits); 3512 3513 if (srcp) 3514 return find_next_bit(srcp, nr_bits, n + 1); 3515 3516 return n + 1; 3517 } 3518 3519 /** 3520 * netif_attrmask_next_and - get the next CPU/Rx queue in *src1p & *src2p 3521 * @n: CPU/Rx queue index 3522 * @src1p: the first CPUs/Rx queues mask pointer 3523 * @src2p: the second CPUs/Rx queues mask pointer 3524 * @nr_bits: number of bits in the bitmask 3525 * 3526 * Returns >= nr_bits if no further CPUs/Rx queues set in both. 3527 */ 3528 static inline int netif_attrmask_next_and(int n, const unsigned long *src1p, 3529 const unsigned long *src2p, 3530 unsigned int nr_bits) 3531 { 3532 /* -1 is a legal arg here. */ 3533 if (n != -1) 3534 cpu_max_bits_warn(n, nr_bits); 3535 3536 if (src1p && src2p) 3537 return find_next_and_bit(src1p, src2p, nr_bits, n + 1); 3538 else if (src1p) 3539 return find_next_bit(src1p, nr_bits, n + 1); 3540 else if (src2p) 3541 return find_next_bit(src2p, nr_bits, n + 1); 3542 3543 return n + 1; 3544 } 3545 #else 3546 static inline int netif_set_xps_queue(struct net_device *dev, 3547 const struct cpumask *mask, 3548 u16 index) 3549 { 3550 return 0; 3551 } 3552 3553 static inline int __netif_set_xps_queue(struct net_device *dev, 3554 const unsigned long *mask, 3555 u16 index, bool is_rxqs_map) 3556 { 3557 return 0; 3558 } 3559 #endif 3560 3561 /** 3562 * netif_is_multiqueue - test if device has multiple transmit queues 3563 * @dev: network device 3564 * 3565 * Check if device has multiple transmit queues 3566 */ 3567 static inline bool netif_is_multiqueue(const struct net_device *dev) 3568 { 3569 return dev->num_tx_queues > 1; 3570 } 3571 3572 int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq); 3573 3574 #ifdef CONFIG_SYSFS 3575 int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq); 3576 #else 3577 static inline int netif_set_real_num_rx_queues(struct net_device *dev, 3578 unsigned int rxqs) 3579 { 3580 dev->real_num_rx_queues = rxqs; 3581 return 0; 3582 } 3583 #endif 3584 3585 static inline struct netdev_rx_queue * 3586 __netif_get_rx_queue(struct net_device *dev, unsigned int rxq) 3587 { 3588 return dev->_rx + rxq; 3589 } 3590 3591 #ifdef CONFIG_SYSFS 3592 static inline unsigned int get_netdev_rx_queue_index( 3593 struct netdev_rx_queue *queue) 3594 { 3595 struct net_device *dev = queue->dev; 3596 int index = queue - dev->_rx; 3597 3598 BUG_ON(index >= dev->num_rx_queues); 3599 return index; 3600 } 3601 #endif 3602 3603 #define DEFAULT_MAX_NUM_RSS_QUEUES (8) 3604 int netif_get_num_default_rss_queues(void); 3605 3606 enum skb_free_reason { 3607 SKB_REASON_CONSUMED, 3608 SKB_REASON_DROPPED, 3609 }; 3610 3611 void __dev_kfree_skb_irq(struct sk_buff *skb, enum skb_free_reason reason); 3612 void __dev_kfree_skb_any(struct sk_buff *skb, enum skb_free_reason reason); 3613 3614 /* 3615 * It is not allowed to call kfree_skb() or consume_skb() from hardware 3616 * interrupt context or with hardware interrupts being disabled. 3617 * (in_irq() || irqs_disabled()) 3618 * 3619 * We provide four helpers that can be used in following contexts : 3620 * 3621 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context, 3622 * replacing kfree_skb(skb) 3623 * 3624 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context. 3625 * Typically used in place of consume_skb(skb) in TX completion path 3626 * 3627 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context, 3628 * replacing kfree_skb(skb) 3629 * 3630 * dev_consume_skb_any(skb) when caller doesn't know its current irq context, 3631 * and consumed a packet. Used in place of consume_skb(skb) 3632 */ 3633 static inline void dev_kfree_skb_irq(struct sk_buff *skb) 3634 { 3635 __dev_kfree_skb_irq(skb, SKB_REASON_DROPPED); 3636 } 3637 3638 static inline void dev_consume_skb_irq(struct sk_buff *skb) 3639 { 3640 __dev_kfree_skb_irq(skb, SKB_REASON_CONSUMED); 3641 } 3642 3643 static inline void dev_kfree_skb_any(struct sk_buff *skb) 3644 { 3645 __dev_kfree_skb_any(skb, SKB_REASON_DROPPED); 3646 } 3647 3648 static inline void dev_consume_skb_any(struct sk_buff *skb) 3649 { 3650 __dev_kfree_skb_any(skb, SKB_REASON_CONSUMED); 3651 } 3652 3653 void generic_xdp_tx(struct sk_buff *skb, struct bpf_prog *xdp_prog); 3654 int do_xdp_generic(struct bpf_prog *xdp_prog, struct sk_buff *skb); 3655 int netif_rx(struct sk_buff *skb); 3656 int netif_rx_ni(struct sk_buff *skb); 3657 int netif_receive_skb(struct sk_buff *skb); 3658 int netif_receive_skb_core(struct sk_buff *skb); 3659 void netif_receive_skb_list(struct list_head *head); 3660 gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb); 3661 void napi_gro_flush(struct napi_struct *napi, bool flush_old); 3662 struct sk_buff *napi_get_frags(struct napi_struct *napi); 3663 gro_result_t napi_gro_frags(struct napi_struct *napi); 3664 struct packet_offload *gro_find_receive_by_type(__be16 type); 3665 struct packet_offload *gro_find_complete_by_type(__be16 type); 3666 3667 static inline void napi_free_frags(struct napi_struct *napi) 3668 { 3669 kfree_skb(napi->skb); 3670 napi->skb = NULL; 3671 } 3672 3673 bool netdev_is_rx_handler_busy(struct net_device *dev); 3674 int netdev_rx_handler_register(struct net_device *dev, 3675 rx_handler_func_t *rx_handler, 3676 void *rx_handler_data); 3677 void netdev_rx_handler_unregister(struct net_device *dev); 3678 3679 bool dev_valid_name(const char *name); 3680 int dev_ioctl(struct net *net, unsigned int cmd, struct ifreq *ifr, 3681 bool *need_copyout); 3682 int dev_ifconf(struct net *net, struct ifconf *, int); 3683 int dev_ethtool(struct net *net, struct ifreq *); 3684 unsigned int dev_get_flags(const struct net_device *); 3685 int __dev_change_flags(struct net_device *dev, unsigned int flags, 3686 struct netlink_ext_ack *extack); 3687 int dev_change_flags(struct net_device *dev, unsigned int flags, 3688 struct netlink_ext_ack *extack); 3689 void __dev_notify_flags(struct net_device *, unsigned int old_flags, 3690 unsigned int gchanges); 3691 int dev_change_name(struct net_device *, const char *); 3692 int dev_set_alias(struct net_device *, const char *, size_t); 3693 int dev_get_alias(const struct net_device *, char *, size_t); 3694 int dev_change_net_namespace(struct net_device *, struct net *, const char *); 3695 int __dev_set_mtu(struct net_device *, int); 3696 int dev_set_mtu_ext(struct net_device *dev, int mtu, 3697 struct netlink_ext_ack *extack); 3698 int dev_set_mtu(struct net_device *, int); 3699 int dev_change_tx_queue_len(struct net_device *, unsigned long); 3700 void dev_set_group(struct net_device *, int); 3701 int dev_pre_changeaddr_notify(struct net_device *dev, const char *addr, 3702 struct netlink_ext_ack *extack); 3703 int dev_set_mac_address(struct net_device *dev, struct sockaddr *sa, 3704 struct netlink_ext_ack *extack); 3705 int dev_change_carrier(struct net_device *, bool new_carrier); 3706 int dev_get_phys_port_id(struct net_device *dev, 3707 struct netdev_phys_item_id *ppid); 3708 int dev_get_phys_port_name(struct net_device *dev, 3709 char *name, size_t len); 3710 int dev_get_port_parent_id(struct net_device *dev, 3711 struct netdev_phys_item_id *ppid, bool recurse); 3712 bool netdev_port_same_parent_id(struct net_device *a, struct net_device *b); 3713 int dev_change_proto_down(struct net_device *dev, bool proto_down); 3714 int dev_change_proto_down_generic(struct net_device *dev, bool proto_down); 3715 struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev, bool *again); 3716 struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev, 3717 struct netdev_queue *txq, int *ret); 3718 3719 typedef int (*bpf_op_t)(struct net_device *dev, struct netdev_bpf *bpf); 3720 int dev_change_xdp_fd(struct net_device *dev, struct netlink_ext_ack *extack, 3721 int fd, u32 flags); 3722 u32 __dev_xdp_query(struct net_device *dev, bpf_op_t xdp_op, 3723 enum bpf_netdev_command cmd); 3724 int xdp_umem_query(struct net_device *dev, u16 queue_id); 3725 3726 int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb); 3727 int dev_forward_skb(struct net_device *dev, struct sk_buff *skb); 3728 bool is_skb_forwardable(const struct net_device *dev, 3729 const struct sk_buff *skb); 3730 3731 static __always_inline int ____dev_forward_skb(struct net_device *dev, 3732 struct sk_buff *skb) 3733 { 3734 if (skb_orphan_frags(skb, GFP_ATOMIC) || 3735 unlikely(!is_skb_forwardable(dev, skb))) { 3736 atomic_long_inc(&dev->rx_dropped); 3737 kfree_skb(skb); 3738 return NET_RX_DROP; 3739 } 3740 3741 skb_scrub_packet(skb, true); 3742 skb->priority = 0; 3743 return 0; 3744 } 3745 3746 bool dev_nit_active(struct net_device *dev); 3747 void dev_queue_xmit_nit(struct sk_buff *skb, struct net_device *dev); 3748 3749 extern int netdev_budget; 3750 extern unsigned int netdev_budget_usecs; 3751 3752 /* Called by rtnetlink.c:rtnl_unlock() */ 3753 void netdev_run_todo(void); 3754 3755 /** 3756 * dev_put - release reference to device 3757 * @dev: network device 3758 * 3759 * Release reference to device to allow it to be freed. 3760 */ 3761 static inline void dev_put(struct net_device *dev) 3762 { 3763 this_cpu_dec(*dev->pcpu_refcnt); 3764 } 3765 3766 /** 3767 * dev_hold - get reference to device 3768 * @dev: network device 3769 * 3770 * Hold reference to device to keep it from being freed. 3771 */ 3772 static inline void dev_hold(struct net_device *dev) 3773 { 3774 this_cpu_inc(*dev->pcpu_refcnt); 3775 } 3776 3777 /* Carrier loss detection, dial on demand. The functions netif_carrier_on 3778 * and _off may be called from IRQ context, but it is caller 3779 * who is responsible for serialization of these calls. 3780 * 3781 * The name carrier is inappropriate, these functions should really be 3782 * called netif_lowerlayer_*() because they represent the state of any 3783 * kind of lower layer not just hardware media. 3784 */ 3785 3786 void linkwatch_init_dev(struct net_device *dev); 3787 void linkwatch_fire_event(struct net_device *dev); 3788 void linkwatch_forget_dev(struct net_device *dev); 3789 3790 /** 3791 * netif_carrier_ok - test if carrier present 3792 * @dev: network device 3793 * 3794 * Check if carrier is present on device 3795 */ 3796 static inline bool netif_carrier_ok(const struct net_device *dev) 3797 { 3798 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state); 3799 } 3800 3801 unsigned long dev_trans_start(struct net_device *dev); 3802 3803 void __netdev_watchdog_up(struct net_device *dev); 3804 3805 void netif_carrier_on(struct net_device *dev); 3806 3807 void netif_carrier_off(struct net_device *dev); 3808 3809 /** 3810 * netif_dormant_on - mark device as dormant. 3811 * @dev: network device 3812 * 3813 * Mark device as dormant (as per RFC2863). 3814 * 3815 * The dormant state indicates that the relevant interface is not 3816 * actually in a condition to pass packets (i.e., it is not 'up') but is 3817 * in a "pending" state, waiting for some external event. For "on- 3818 * demand" interfaces, this new state identifies the situation where the 3819 * interface is waiting for events to place it in the up state. 3820 */ 3821 static inline void netif_dormant_on(struct net_device *dev) 3822 { 3823 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state)) 3824 linkwatch_fire_event(dev); 3825 } 3826 3827 /** 3828 * netif_dormant_off - set device as not dormant. 3829 * @dev: network device 3830 * 3831 * Device is not in dormant state. 3832 */ 3833 static inline void netif_dormant_off(struct net_device *dev) 3834 { 3835 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state)) 3836 linkwatch_fire_event(dev); 3837 } 3838 3839 /** 3840 * netif_dormant - test if device is dormant 3841 * @dev: network device 3842 * 3843 * Check if device is dormant. 3844 */ 3845 static inline bool netif_dormant(const struct net_device *dev) 3846 { 3847 return test_bit(__LINK_STATE_DORMANT, &dev->state); 3848 } 3849 3850 3851 /** 3852 * netif_oper_up - test if device is operational 3853 * @dev: network device 3854 * 3855 * Check if carrier is operational 3856 */ 3857 static inline bool netif_oper_up(const struct net_device *dev) 3858 { 3859 return (dev->operstate == IF_OPER_UP || 3860 dev->operstate == IF_OPER_UNKNOWN /* backward compat */); 3861 } 3862 3863 /** 3864 * netif_device_present - is device available or removed 3865 * @dev: network device 3866 * 3867 * Check if device has not been removed from system. 3868 */ 3869 static inline bool netif_device_present(struct net_device *dev) 3870 { 3871 return test_bit(__LINK_STATE_PRESENT, &dev->state); 3872 } 3873 3874 void netif_device_detach(struct net_device *dev); 3875 3876 void netif_device_attach(struct net_device *dev); 3877 3878 /* 3879 * Network interface message level settings 3880 */ 3881 3882 enum { 3883 NETIF_MSG_DRV = 0x0001, 3884 NETIF_MSG_PROBE = 0x0002, 3885 NETIF_MSG_LINK = 0x0004, 3886 NETIF_MSG_TIMER = 0x0008, 3887 NETIF_MSG_IFDOWN = 0x0010, 3888 NETIF_MSG_IFUP = 0x0020, 3889 NETIF_MSG_RX_ERR = 0x0040, 3890 NETIF_MSG_TX_ERR = 0x0080, 3891 NETIF_MSG_TX_QUEUED = 0x0100, 3892 NETIF_MSG_INTR = 0x0200, 3893 NETIF_MSG_TX_DONE = 0x0400, 3894 NETIF_MSG_RX_STATUS = 0x0800, 3895 NETIF_MSG_PKTDATA = 0x1000, 3896 NETIF_MSG_HW = 0x2000, 3897 NETIF_MSG_WOL = 0x4000, 3898 }; 3899 3900 #define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV) 3901 #define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE) 3902 #define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK) 3903 #define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER) 3904 #define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN) 3905 #define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP) 3906 #define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR) 3907 #define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR) 3908 #define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED) 3909 #define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR) 3910 #define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE) 3911 #define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS) 3912 #define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA) 3913 #define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW) 3914 #define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL) 3915 3916 static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits) 3917 { 3918 /* use default */ 3919 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8)) 3920 return default_msg_enable_bits; 3921 if (debug_value == 0) /* no output */ 3922 return 0; 3923 /* set low N bits */ 3924 return (1U << debug_value) - 1; 3925 } 3926 3927 static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu) 3928 { 3929 spin_lock(&txq->_xmit_lock); 3930 txq->xmit_lock_owner = cpu; 3931 } 3932 3933 static inline bool __netif_tx_acquire(struct netdev_queue *txq) 3934 { 3935 __acquire(&txq->_xmit_lock); 3936 return true; 3937 } 3938 3939 static inline void __netif_tx_release(struct netdev_queue *txq) 3940 { 3941 __release(&txq->_xmit_lock); 3942 } 3943 3944 static inline void __netif_tx_lock_bh(struct netdev_queue *txq) 3945 { 3946 spin_lock_bh(&txq->_xmit_lock); 3947 txq->xmit_lock_owner = smp_processor_id(); 3948 } 3949 3950 static inline bool __netif_tx_trylock(struct netdev_queue *txq) 3951 { 3952 bool ok = spin_trylock(&txq->_xmit_lock); 3953 if (likely(ok)) 3954 txq->xmit_lock_owner = smp_processor_id(); 3955 return ok; 3956 } 3957 3958 static inline void __netif_tx_unlock(struct netdev_queue *txq) 3959 { 3960 txq->xmit_lock_owner = -1; 3961 spin_unlock(&txq->_xmit_lock); 3962 } 3963 3964 static inline void __netif_tx_unlock_bh(struct netdev_queue *txq) 3965 { 3966 txq->xmit_lock_owner = -1; 3967 spin_unlock_bh(&txq->_xmit_lock); 3968 } 3969 3970 static inline void txq_trans_update(struct netdev_queue *txq) 3971 { 3972 if (txq->xmit_lock_owner != -1) 3973 txq->trans_start = jiffies; 3974 } 3975 3976 /* legacy drivers only, netdev_start_xmit() sets txq->trans_start */ 3977 static inline void netif_trans_update(struct net_device *dev) 3978 { 3979 struct netdev_queue *txq = netdev_get_tx_queue(dev, 0); 3980 3981 if (txq->trans_start != jiffies) 3982 txq->trans_start = jiffies; 3983 } 3984 3985 /** 3986 * netif_tx_lock - grab network device transmit lock 3987 * @dev: network device 3988 * 3989 * Get network device transmit lock 3990 */ 3991 static inline void netif_tx_lock(struct net_device *dev) 3992 { 3993 unsigned int i; 3994 int cpu; 3995 3996 spin_lock(&dev->tx_global_lock); 3997 cpu = smp_processor_id(); 3998 for (i = 0; i < dev->num_tx_queues; i++) { 3999 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 4000 4001 /* We are the only thread of execution doing a 4002 * freeze, but we have to grab the _xmit_lock in 4003 * order to synchronize with threads which are in 4004 * the ->hard_start_xmit() handler and already 4005 * checked the frozen bit. 4006 */ 4007 __netif_tx_lock(txq, cpu); 4008 set_bit(__QUEUE_STATE_FROZEN, &txq->state); 4009 __netif_tx_unlock(txq); 4010 } 4011 } 4012 4013 static inline void netif_tx_lock_bh(struct net_device *dev) 4014 { 4015 local_bh_disable(); 4016 netif_tx_lock(dev); 4017 } 4018 4019 static inline void netif_tx_unlock(struct net_device *dev) 4020 { 4021 unsigned int i; 4022 4023 for (i = 0; i < dev->num_tx_queues; i++) { 4024 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 4025 4026 /* No need to grab the _xmit_lock here. If the 4027 * queue is not stopped for another reason, we 4028 * force a schedule. 4029 */ 4030 clear_bit(__QUEUE_STATE_FROZEN, &txq->state); 4031 netif_schedule_queue(txq); 4032 } 4033 spin_unlock(&dev->tx_global_lock); 4034 } 4035 4036 static inline void netif_tx_unlock_bh(struct net_device *dev) 4037 { 4038 netif_tx_unlock(dev); 4039 local_bh_enable(); 4040 } 4041 4042 #define HARD_TX_LOCK(dev, txq, cpu) { \ 4043 if ((dev->features & NETIF_F_LLTX) == 0) { \ 4044 __netif_tx_lock(txq, cpu); \ 4045 } else { \ 4046 __netif_tx_acquire(txq); \ 4047 } \ 4048 } 4049 4050 #define HARD_TX_TRYLOCK(dev, txq) \ 4051 (((dev->features & NETIF_F_LLTX) == 0) ? \ 4052 __netif_tx_trylock(txq) : \ 4053 __netif_tx_acquire(txq)) 4054 4055 #define HARD_TX_UNLOCK(dev, txq) { \ 4056 if ((dev->features & NETIF_F_LLTX) == 0) { \ 4057 __netif_tx_unlock(txq); \ 4058 } else { \ 4059 __netif_tx_release(txq); \ 4060 } \ 4061 } 4062 4063 static inline void netif_tx_disable(struct net_device *dev) 4064 { 4065 unsigned int i; 4066 int cpu; 4067 4068 local_bh_disable(); 4069 cpu = smp_processor_id(); 4070 for (i = 0; i < dev->num_tx_queues; i++) { 4071 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 4072 4073 __netif_tx_lock(txq, cpu); 4074 netif_tx_stop_queue(txq); 4075 __netif_tx_unlock(txq); 4076 } 4077 local_bh_enable(); 4078 } 4079 4080 static inline void netif_addr_lock(struct net_device *dev) 4081 { 4082 spin_lock(&dev->addr_list_lock); 4083 } 4084 4085 static inline void netif_addr_lock_bh(struct net_device *dev) 4086 { 4087 spin_lock_bh(&dev->addr_list_lock); 4088 } 4089 4090 static inline void netif_addr_unlock(struct net_device *dev) 4091 { 4092 spin_unlock(&dev->addr_list_lock); 4093 } 4094 4095 static inline void netif_addr_unlock_bh(struct net_device *dev) 4096 { 4097 spin_unlock_bh(&dev->addr_list_lock); 4098 } 4099 4100 /* 4101 * dev_addrs walker. Should be used only for read access. Call with 4102 * rcu_read_lock held. 4103 */ 4104 #define for_each_dev_addr(dev, ha) \ 4105 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list) 4106 4107 /* These functions live elsewhere (drivers/net/net_init.c, but related) */ 4108 4109 void ether_setup(struct net_device *dev); 4110 4111 /* Support for loadable net-drivers */ 4112 struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name, 4113 unsigned char name_assign_type, 4114 void (*setup)(struct net_device *), 4115 unsigned int txqs, unsigned int rxqs); 4116 #define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \ 4117 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1) 4118 4119 #define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \ 4120 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \ 4121 count) 4122 4123 int register_netdev(struct net_device *dev); 4124 void unregister_netdev(struct net_device *dev); 4125 4126 /* General hardware address lists handling functions */ 4127 int __hw_addr_sync(struct netdev_hw_addr_list *to_list, 4128 struct netdev_hw_addr_list *from_list, int addr_len); 4129 void __hw_addr_unsync(struct netdev_hw_addr_list *to_list, 4130 struct netdev_hw_addr_list *from_list, int addr_len); 4131 int __hw_addr_sync_dev(struct netdev_hw_addr_list *list, 4132 struct net_device *dev, 4133 int (*sync)(struct net_device *, const unsigned char *), 4134 int (*unsync)(struct net_device *, 4135 const unsigned char *)); 4136 int __hw_addr_ref_sync_dev(struct netdev_hw_addr_list *list, 4137 struct net_device *dev, 4138 int (*sync)(struct net_device *, 4139 const unsigned char *, int), 4140 int (*unsync)(struct net_device *, 4141 const unsigned char *, int)); 4142 void __hw_addr_ref_unsync_dev(struct netdev_hw_addr_list *list, 4143 struct net_device *dev, 4144 int (*unsync)(struct net_device *, 4145 const unsigned char *, int)); 4146 void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list, 4147 struct net_device *dev, 4148 int (*unsync)(struct net_device *, 4149 const unsigned char *)); 4150 void __hw_addr_init(struct netdev_hw_addr_list *list); 4151 4152 /* Functions used for device addresses handling */ 4153 int dev_addr_add(struct net_device *dev, const unsigned char *addr, 4154 unsigned char addr_type); 4155 int dev_addr_del(struct net_device *dev, const unsigned char *addr, 4156 unsigned char addr_type); 4157 void dev_addr_flush(struct net_device *dev); 4158 int dev_addr_init(struct net_device *dev); 4159 4160 /* Functions used for unicast addresses handling */ 4161 int dev_uc_add(struct net_device *dev, const unsigned char *addr); 4162 int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr); 4163 int dev_uc_del(struct net_device *dev, const unsigned char *addr); 4164 int dev_uc_sync(struct net_device *to, struct net_device *from); 4165 int dev_uc_sync_multiple(struct net_device *to, struct net_device *from); 4166 void dev_uc_unsync(struct net_device *to, struct net_device *from); 4167 void dev_uc_flush(struct net_device *dev); 4168 void dev_uc_init(struct net_device *dev); 4169 4170 /** 4171 * __dev_uc_sync - Synchonize device's unicast list 4172 * @dev: device to sync 4173 * @sync: function to call if address should be added 4174 * @unsync: function to call if address should be removed 4175 * 4176 * Add newly added addresses to the interface, and release 4177 * addresses that have been deleted. 4178 */ 4179 static inline int __dev_uc_sync(struct net_device *dev, 4180 int (*sync)(struct net_device *, 4181 const unsigned char *), 4182 int (*unsync)(struct net_device *, 4183 const unsigned char *)) 4184 { 4185 return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync); 4186 } 4187 4188 /** 4189 * __dev_uc_unsync - Remove synchronized addresses from device 4190 * @dev: device to sync 4191 * @unsync: function to call if address should be removed 4192 * 4193 * Remove all addresses that were added to the device by dev_uc_sync(). 4194 */ 4195 static inline void __dev_uc_unsync(struct net_device *dev, 4196 int (*unsync)(struct net_device *, 4197 const unsigned char *)) 4198 { 4199 __hw_addr_unsync_dev(&dev->uc, dev, unsync); 4200 } 4201 4202 /* Functions used for multicast addresses handling */ 4203 int dev_mc_add(struct net_device *dev, const unsigned char *addr); 4204 int dev_mc_add_global(struct net_device *dev, const unsigned char *addr); 4205 int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr); 4206 int dev_mc_del(struct net_device *dev, const unsigned char *addr); 4207 int dev_mc_del_global(struct net_device *dev, const unsigned char *addr); 4208 int dev_mc_sync(struct net_device *to, struct net_device *from); 4209 int dev_mc_sync_multiple(struct net_device *to, struct net_device *from); 4210 void dev_mc_unsync(struct net_device *to, struct net_device *from); 4211 void dev_mc_flush(struct net_device *dev); 4212 void dev_mc_init(struct net_device *dev); 4213 4214 /** 4215 * __dev_mc_sync - Synchonize device's multicast list 4216 * @dev: device to sync 4217 * @sync: function to call if address should be added 4218 * @unsync: function to call if address should be removed 4219 * 4220 * Add newly added addresses to the interface, and release 4221 * addresses that have been deleted. 4222 */ 4223 static inline int __dev_mc_sync(struct net_device *dev, 4224 int (*sync)(struct net_device *, 4225 const unsigned char *), 4226 int (*unsync)(struct net_device *, 4227 const unsigned char *)) 4228 { 4229 return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync); 4230 } 4231 4232 /** 4233 * __dev_mc_unsync - Remove synchronized addresses from device 4234 * @dev: device to sync 4235 * @unsync: function to call if address should be removed 4236 * 4237 * Remove all addresses that were added to the device by dev_mc_sync(). 4238 */ 4239 static inline void __dev_mc_unsync(struct net_device *dev, 4240 int (*unsync)(struct net_device *, 4241 const unsigned char *)) 4242 { 4243 __hw_addr_unsync_dev(&dev->mc, dev, unsync); 4244 } 4245 4246 /* Functions used for secondary unicast and multicast support */ 4247 void dev_set_rx_mode(struct net_device *dev); 4248 void __dev_set_rx_mode(struct net_device *dev); 4249 int dev_set_promiscuity(struct net_device *dev, int inc); 4250 int dev_set_allmulti(struct net_device *dev, int inc); 4251 void netdev_state_change(struct net_device *dev); 4252 void netdev_notify_peers(struct net_device *dev); 4253 void netdev_features_change(struct net_device *dev); 4254 /* Load a device via the kmod */ 4255 void dev_load(struct net *net, const char *name); 4256 struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev, 4257 struct rtnl_link_stats64 *storage); 4258 void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64, 4259 const struct net_device_stats *netdev_stats); 4260 4261 extern int netdev_max_backlog; 4262 extern int netdev_tstamp_prequeue; 4263 extern int weight_p; 4264 extern int dev_weight_rx_bias; 4265 extern int dev_weight_tx_bias; 4266 extern int dev_rx_weight; 4267 extern int dev_tx_weight; 4268 extern int gro_normal_batch; 4269 4270 bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev); 4271 struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev, 4272 struct list_head **iter); 4273 struct net_device *netdev_all_upper_get_next_dev_rcu(struct net_device *dev, 4274 struct list_head **iter); 4275 4276 /* iterate through upper list, must be called under RCU read lock */ 4277 #define netdev_for_each_upper_dev_rcu(dev, updev, iter) \ 4278 for (iter = &(dev)->adj_list.upper, \ 4279 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \ 4280 updev; \ 4281 updev = netdev_upper_get_next_dev_rcu(dev, &(iter))) 4282 4283 int netdev_walk_all_upper_dev_rcu(struct net_device *dev, 4284 int (*fn)(struct net_device *upper_dev, 4285 void *data), 4286 void *data); 4287 4288 bool netdev_has_upper_dev_all_rcu(struct net_device *dev, 4289 struct net_device *upper_dev); 4290 4291 bool netdev_has_any_upper_dev(struct net_device *dev); 4292 4293 void *netdev_lower_get_next_private(struct net_device *dev, 4294 struct list_head **iter); 4295 void *netdev_lower_get_next_private_rcu(struct net_device *dev, 4296 struct list_head **iter); 4297 4298 #define netdev_for_each_lower_private(dev, priv, iter) \ 4299 for (iter = (dev)->adj_list.lower.next, \ 4300 priv = netdev_lower_get_next_private(dev, &(iter)); \ 4301 priv; \ 4302 priv = netdev_lower_get_next_private(dev, &(iter))) 4303 4304 #define netdev_for_each_lower_private_rcu(dev, priv, iter) \ 4305 for (iter = &(dev)->adj_list.lower, \ 4306 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \ 4307 priv; \ 4308 priv = netdev_lower_get_next_private_rcu(dev, &(iter))) 4309 4310 void *netdev_lower_get_next(struct net_device *dev, 4311 struct list_head **iter); 4312 4313 #define netdev_for_each_lower_dev(dev, ldev, iter) \ 4314 for (iter = (dev)->adj_list.lower.next, \ 4315 ldev = netdev_lower_get_next(dev, &(iter)); \ 4316 ldev; \ 4317 ldev = netdev_lower_get_next(dev, &(iter))) 4318 4319 struct net_device *netdev_all_lower_get_next(struct net_device *dev, 4320 struct list_head **iter); 4321 struct net_device *netdev_all_lower_get_next_rcu(struct net_device *dev, 4322 struct list_head **iter); 4323 4324 int netdev_walk_all_lower_dev(struct net_device *dev, 4325 int (*fn)(struct net_device *lower_dev, 4326 void *data), 4327 void *data); 4328 int netdev_walk_all_lower_dev_rcu(struct net_device *dev, 4329 int (*fn)(struct net_device *lower_dev, 4330 void *data), 4331 void *data); 4332 4333 void *netdev_adjacent_get_private(struct list_head *adj_list); 4334 void *netdev_lower_get_first_private_rcu(struct net_device *dev); 4335 struct net_device *netdev_master_upper_dev_get(struct net_device *dev); 4336 struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev); 4337 int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev, 4338 struct netlink_ext_ack *extack); 4339 int netdev_master_upper_dev_link(struct net_device *dev, 4340 struct net_device *upper_dev, 4341 void *upper_priv, void *upper_info, 4342 struct netlink_ext_ack *extack); 4343 void netdev_upper_dev_unlink(struct net_device *dev, 4344 struct net_device *upper_dev); 4345 int netdev_adjacent_change_prepare(struct net_device *old_dev, 4346 struct net_device *new_dev, 4347 struct net_device *dev, 4348 struct netlink_ext_ack *extack); 4349 void netdev_adjacent_change_commit(struct net_device *old_dev, 4350 struct net_device *new_dev, 4351 struct net_device *dev); 4352 void netdev_adjacent_change_abort(struct net_device *old_dev, 4353 struct net_device *new_dev, 4354 struct net_device *dev); 4355 void netdev_adjacent_rename_links(struct net_device *dev, char *oldname); 4356 void *netdev_lower_dev_get_private(struct net_device *dev, 4357 struct net_device *lower_dev); 4358 void netdev_lower_state_changed(struct net_device *lower_dev, 4359 void *lower_state_info); 4360 4361 /* RSS keys are 40 or 52 bytes long */ 4362 #define NETDEV_RSS_KEY_LEN 52 4363 extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN] __read_mostly; 4364 void netdev_rss_key_fill(void *buffer, size_t len); 4365 4366 int skb_checksum_help(struct sk_buff *skb); 4367 int skb_crc32c_csum_help(struct sk_buff *skb); 4368 int skb_csum_hwoffload_help(struct sk_buff *skb, 4369 const netdev_features_t features); 4370 4371 struct sk_buff *__skb_gso_segment(struct sk_buff *skb, 4372 netdev_features_t features, bool tx_path); 4373 struct sk_buff *skb_mac_gso_segment(struct sk_buff *skb, 4374 netdev_features_t features); 4375 4376 struct netdev_bonding_info { 4377 ifslave slave; 4378 ifbond master; 4379 }; 4380 4381 struct netdev_notifier_bonding_info { 4382 struct netdev_notifier_info info; /* must be first */ 4383 struct netdev_bonding_info bonding_info; 4384 }; 4385 4386 void netdev_bonding_info_change(struct net_device *dev, 4387 struct netdev_bonding_info *bonding_info); 4388 4389 static inline 4390 struct sk_buff *skb_gso_segment(struct sk_buff *skb, netdev_features_t features) 4391 { 4392 return __skb_gso_segment(skb, features, true); 4393 } 4394 __be16 skb_network_protocol(struct sk_buff *skb, int *depth); 4395 4396 static inline bool can_checksum_protocol(netdev_features_t features, 4397 __be16 protocol) 4398 { 4399 if (protocol == htons(ETH_P_FCOE)) 4400 return !!(features & NETIF_F_FCOE_CRC); 4401 4402 /* Assume this is an IP checksum (not SCTP CRC) */ 4403 4404 if (features & NETIF_F_HW_CSUM) { 4405 /* Can checksum everything */ 4406 return true; 4407 } 4408 4409 switch (protocol) { 4410 case htons(ETH_P_IP): 4411 return !!(features & NETIF_F_IP_CSUM); 4412 case htons(ETH_P_IPV6): 4413 return !!(features & NETIF_F_IPV6_CSUM); 4414 default: 4415 return false; 4416 } 4417 } 4418 4419 #ifdef CONFIG_BUG 4420 void netdev_rx_csum_fault(struct net_device *dev, struct sk_buff *skb); 4421 #else 4422 static inline void netdev_rx_csum_fault(struct net_device *dev, 4423 struct sk_buff *skb) 4424 { 4425 } 4426 #endif 4427 /* rx skb timestamps */ 4428 void net_enable_timestamp(void); 4429 void net_disable_timestamp(void); 4430 4431 #ifdef CONFIG_PROC_FS 4432 int __init dev_proc_init(void); 4433 #else 4434 #define dev_proc_init() 0 4435 #endif 4436 4437 static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops, 4438 struct sk_buff *skb, struct net_device *dev, 4439 bool more) 4440 { 4441 __this_cpu_write(softnet_data.xmit.more, more); 4442 return ops->ndo_start_xmit(skb, dev); 4443 } 4444 4445 static inline bool netdev_xmit_more(void) 4446 { 4447 return __this_cpu_read(softnet_data.xmit.more); 4448 } 4449 4450 static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev, 4451 struct netdev_queue *txq, bool more) 4452 { 4453 const struct net_device_ops *ops = dev->netdev_ops; 4454 netdev_tx_t rc; 4455 4456 rc = __netdev_start_xmit(ops, skb, dev, more); 4457 if (rc == NETDEV_TX_OK) 4458 txq_trans_update(txq); 4459 4460 return rc; 4461 } 4462 4463 int netdev_class_create_file_ns(const struct class_attribute *class_attr, 4464 const void *ns); 4465 void netdev_class_remove_file_ns(const struct class_attribute *class_attr, 4466 const void *ns); 4467 4468 static inline int netdev_class_create_file(const struct class_attribute *class_attr) 4469 { 4470 return netdev_class_create_file_ns(class_attr, NULL); 4471 } 4472 4473 static inline void netdev_class_remove_file(const struct class_attribute *class_attr) 4474 { 4475 netdev_class_remove_file_ns(class_attr, NULL); 4476 } 4477 4478 extern const struct kobj_ns_type_operations net_ns_type_operations; 4479 4480 const char *netdev_drivername(const struct net_device *dev); 4481 4482 void linkwatch_run_queue(void); 4483 4484 static inline netdev_features_t netdev_intersect_features(netdev_features_t f1, 4485 netdev_features_t f2) 4486 { 4487 if ((f1 ^ f2) & NETIF_F_HW_CSUM) { 4488 if (f1 & NETIF_F_HW_CSUM) 4489 f1 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM); 4490 else 4491 f2 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM); 4492 } 4493 4494 return f1 & f2; 4495 } 4496 4497 static inline netdev_features_t netdev_get_wanted_features( 4498 struct net_device *dev) 4499 { 4500 return (dev->features & ~dev->hw_features) | dev->wanted_features; 4501 } 4502 netdev_features_t netdev_increment_features(netdev_features_t all, 4503 netdev_features_t one, netdev_features_t mask); 4504 4505 /* Allow TSO being used on stacked device : 4506 * Performing the GSO segmentation before last device 4507 * is a performance improvement. 4508 */ 4509 static inline netdev_features_t netdev_add_tso_features(netdev_features_t features, 4510 netdev_features_t mask) 4511 { 4512 return netdev_increment_features(features, NETIF_F_ALL_TSO, mask); 4513 } 4514 4515 int __netdev_update_features(struct net_device *dev); 4516 void netdev_update_features(struct net_device *dev); 4517 void netdev_change_features(struct net_device *dev); 4518 4519 void netif_stacked_transfer_operstate(const struct net_device *rootdev, 4520 struct net_device *dev); 4521 4522 netdev_features_t passthru_features_check(struct sk_buff *skb, 4523 struct net_device *dev, 4524 netdev_features_t features); 4525 netdev_features_t netif_skb_features(struct sk_buff *skb); 4526 4527 static inline bool net_gso_ok(netdev_features_t features, int gso_type) 4528 { 4529 netdev_features_t feature = (netdev_features_t)gso_type << NETIF_F_GSO_SHIFT; 4530 4531 /* check flags correspondence */ 4532 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT)); 4533 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT)); 4534 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT)); 4535 BUILD_BUG_ON(SKB_GSO_TCP_FIXEDID != (NETIF_F_TSO_MANGLEID >> NETIF_F_GSO_SHIFT)); 4536 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT)); 4537 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT)); 4538 BUILD_BUG_ON(SKB_GSO_GRE != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT)); 4539 BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT)); 4540 BUILD_BUG_ON(SKB_GSO_IPXIP4 != (NETIF_F_GSO_IPXIP4 >> NETIF_F_GSO_SHIFT)); 4541 BUILD_BUG_ON(SKB_GSO_IPXIP6 != (NETIF_F_GSO_IPXIP6 >> NETIF_F_GSO_SHIFT)); 4542 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT)); 4543 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT)); 4544 BUILD_BUG_ON(SKB_GSO_PARTIAL != (NETIF_F_GSO_PARTIAL >> NETIF_F_GSO_SHIFT)); 4545 BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT)); 4546 BUILD_BUG_ON(SKB_GSO_SCTP != (NETIF_F_GSO_SCTP >> NETIF_F_GSO_SHIFT)); 4547 BUILD_BUG_ON(SKB_GSO_ESP != (NETIF_F_GSO_ESP >> NETIF_F_GSO_SHIFT)); 4548 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_GSO_UDP >> NETIF_F_GSO_SHIFT)); 4549 BUILD_BUG_ON(SKB_GSO_UDP_L4 != (NETIF_F_GSO_UDP_L4 >> NETIF_F_GSO_SHIFT)); 4550 4551 return (features & feature) == feature; 4552 } 4553 4554 static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features) 4555 { 4556 return net_gso_ok(features, skb_shinfo(skb)->gso_type) && 4557 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST)); 4558 } 4559 4560 static inline bool netif_needs_gso(struct sk_buff *skb, 4561 netdev_features_t features) 4562 { 4563 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) || 4564 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) && 4565 (skb->ip_summed != CHECKSUM_UNNECESSARY))); 4566 } 4567 4568 static inline void netif_set_gso_max_size(struct net_device *dev, 4569 unsigned int size) 4570 { 4571 dev->gso_max_size = size; 4572 } 4573 4574 static inline void skb_gso_error_unwind(struct sk_buff *skb, __be16 protocol, 4575 int pulled_hlen, u16 mac_offset, 4576 int mac_len) 4577 { 4578 skb->protocol = protocol; 4579 skb->encapsulation = 1; 4580 skb_push(skb, pulled_hlen); 4581 skb_reset_transport_header(skb); 4582 skb->mac_header = mac_offset; 4583 skb->network_header = skb->mac_header + mac_len; 4584 skb->mac_len = mac_len; 4585 } 4586 4587 static inline bool netif_is_macsec(const struct net_device *dev) 4588 { 4589 return dev->priv_flags & IFF_MACSEC; 4590 } 4591 4592 static inline bool netif_is_macvlan(const struct net_device *dev) 4593 { 4594 return dev->priv_flags & IFF_MACVLAN; 4595 } 4596 4597 static inline bool netif_is_macvlan_port(const struct net_device *dev) 4598 { 4599 return dev->priv_flags & IFF_MACVLAN_PORT; 4600 } 4601 4602 static inline bool netif_is_bond_master(const struct net_device *dev) 4603 { 4604 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING; 4605 } 4606 4607 static inline bool netif_is_bond_slave(const struct net_device *dev) 4608 { 4609 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING; 4610 } 4611 4612 static inline bool netif_supports_nofcs(struct net_device *dev) 4613 { 4614 return dev->priv_flags & IFF_SUPP_NOFCS; 4615 } 4616 4617 static inline bool netif_has_l3_rx_handler(const struct net_device *dev) 4618 { 4619 return dev->priv_flags & IFF_L3MDEV_RX_HANDLER; 4620 } 4621 4622 static inline bool netif_is_l3_master(const struct net_device *dev) 4623 { 4624 return dev->priv_flags & IFF_L3MDEV_MASTER; 4625 } 4626 4627 static inline bool netif_is_l3_slave(const struct net_device *dev) 4628 { 4629 return dev->priv_flags & IFF_L3MDEV_SLAVE; 4630 } 4631 4632 static inline bool netif_is_bridge_master(const struct net_device *dev) 4633 { 4634 return dev->priv_flags & IFF_EBRIDGE; 4635 } 4636 4637 static inline bool netif_is_bridge_port(const struct net_device *dev) 4638 { 4639 return dev->priv_flags & IFF_BRIDGE_PORT; 4640 } 4641 4642 static inline bool netif_is_ovs_master(const struct net_device *dev) 4643 { 4644 return dev->priv_flags & IFF_OPENVSWITCH; 4645 } 4646 4647 static inline bool netif_is_ovs_port(const struct net_device *dev) 4648 { 4649 return dev->priv_flags & IFF_OVS_DATAPATH; 4650 } 4651 4652 static inline bool netif_is_team_master(const struct net_device *dev) 4653 { 4654 return dev->priv_flags & IFF_TEAM; 4655 } 4656 4657 static inline bool netif_is_team_port(const struct net_device *dev) 4658 { 4659 return dev->priv_flags & IFF_TEAM_PORT; 4660 } 4661 4662 static inline bool netif_is_lag_master(const struct net_device *dev) 4663 { 4664 return netif_is_bond_master(dev) || netif_is_team_master(dev); 4665 } 4666 4667 static inline bool netif_is_lag_port(const struct net_device *dev) 4668 { 4669 return netif_is_bond_slave(dev) || netif_is_team_port(dev); 4670 } 4671 4672 static inline bool netif_is_rxfh_configured(const struct net_device *dev) 4673 { 4674 return dev->priv_flags & IFF_RXFH_CONFIGURED; 4675 } 4676 4677 static inline bool netif_is_failover(const struct net_device *dev) 4678 { 4679 return dev->priv_flags & IFF_FAILOVER; 4680 } 4681 4682 static inline bool netif_is_failover_slave(const struct net_device *dev) 4683 { 4684 return dev->priv_flags & IFF_FAILOVER_SLAVE; 4685 } 4686 4687 /* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */ 4688 static inline void netif_keep_dst(struct net_device *dev) 4689 { 4690 dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM); 4691 } 4692 4693 /* return true if dev can't cope with mtu frames that need vlan tag insertion */ 4694 static inline bool netif_reduces_vlan_mtu(struct net_device *dev) 4695 { 4696 /* TODO: reserve and use an additional IFF bit, if we get more users */ 4697 return dev->priv_flags & IFF_MACSEC; 4698 } 4699 4700 extern struct pernet_operations __net_initdata loopback_net_ops; 4701 4702 /* Logging, debugging and troubleshooting/diagnostic helpers. */ 4703 4704 /* netdev_printk helpers, similar to dev_printk */ 4705 4706 static inline const char *netdev_name(const struct net_device *dev) 4707 { 4708 if (!dev->name[0] || strchr(dev->name, '%')) 4709 return "(unnamed net_device)"; 4710 return dev->name; 4711 } 4712 4713 static inline bool netdev_unregistering(const struct net_device *dev) 4714 { 4715 return dev->reg_state == NETREG_UNREGISTERING; 4716 } 4717 4718 static inline const char *netdev_reg_state(const struct net_device *dev) 4719 { 4720 switch (dev->reg_state) { 4721 case NETREG_UNINITIALIZED: return " (uninitialized)"; 4722 case NETREG_REGISTERED: return ""; 4723 case NETREG_UNREGISTERING: return " (unregistering)"; 4724 case NETREG_UNREGISTERED: return " (unregistered)"; 4725 case NETREG_RELEASED: return " (released)"; 4726 case NETREG_DUMMY: return " (dummy)"; 4727 } 4728 4729 WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, dev->reg_state); 4730 return " (unknown)"; 4731 } 4732 4733 __printf(3, 4) __cold 4734 void netdev_printk(const char *level, const struct net_device *dev, 4735 const char *format, ...); 4736 __printf(2, 3) __cold 4737 void netdev_emerg(const struct net_device *dev, const char *format, ...); 4738 __printf(2, 3) __cold 4739 void netdev_alert(const struct net_device *dev, const char *format, ...); 4740 __printf(2, 3) __cold 4741 void netdev_crit(const struct net_device *dev, const char *format, ...); 4742 __printf(2, 3) __cold 4743 void netdev_err(const struct net_device *dev, const char *format, ...); 4744 __printf(2, 3) __cold 4745 void netdev_warn(const struct net_device *dev, const char *format, ...); 4746 __printf(2, 3) __cold 4747 void netdev_notice(const struct net_device *dev, const char *format, ...); 4748 __printf(2, 3) __cold 4749 void netdev_info(const struct net_device *dev, const char *format, ...); 4750 4751 #define netdev_level_once(level, dev, fmt, ...) \ 4752 do { \ 4753 static bool __print_once __read_mostly; \ 4754 \ 4755 if (!__print_once) { \ 4756 __print_once = true; \ 4757 netdev_printk(level, dev, fmt, ##__VA_ARGS__); \ 4758 } \ 4759 } while (0) 4760 4761 #define netdev_emerg_once(dev, fmt, ...) \ 4762 netdev_level_once(KERN_EMERG, dev, fmt, ##__VA_ARGS__) 4763 #define netdev_alert_once(dev, fmt, ...) \ 4764 netdev_level_once(KERN_ALERT, dev, fmt, ##__VA_ARGS__) 4765 #define netdev_crit_once(dev, fmt, ...) \ 4766 netdev_level_once(KERN_CRIT, dev, fmt, ##__VA_ARGS__) 4767 #define netdev_err_once(dev, fmt, ...) \ 4768 netdev_level_once(KERN_ERR, dev, fmt, ##__VA_ARGS__) 4769 #define netdev_warn_once(dev, fmt, ...) \ 4770 netdev_level_once(KERN_WARNING, dev, fmt, ##__VA_ARGS__) 4771 #define netdev_notice_once(dev, fmt, ...) \ 4772 netdev_level_once(KERN_NOTICE, dev, fmt, ##__VA_ARGS__) 4773 #define netdev_info_once(dev, fmt, ...) \ 4774 netdev_level_once(KERN_INFO, dev, fmt, ##__VA_ARGS__) 4775 4776 #define MODULE_ALIAS_NETDEV(device) \ 4777 MODULE_ALIAS("netdev-" device) 4778 4779 #if defined(CONFIG_DYNAMIC_DEBUG) 4780 #define netdev_dbg(__dev, format, args...) \ 4781 do { \ 4782 dynamic_netdev_dbg(__dev, format, ##args); \ 4783 } while (0) 4784 #elif defined(DEBUG) 4785 #define netdev_dbg(__dev, format, args...) \ 4786 netdev_printk(KERN_DEBUG, __dev, format, ##args) 4787 #else 4788 #define netdev_dbg(__dev, format, args...) \ 4789 ({ \ 4790 if (0) \ 4791 netdev_printk(KERN_DEBUG, __dev, format, ##args); \ 4792 }) 4793 #endif 4794 4795 #if defined(VERBOSE_DEBUG) 4796 #define netdev_vdbg netdev_dbg 4797 #else 4798 4799 #define netdev_vdbg(dev, format, args...) \ 4800 ({ \ 4801 if (0) \ 4802 netdev_printk(KERN_DEBUG, dev, format, ##args); \ 4803 0; \ 4804 }) 4805 #endif 4806 4807 /* 4808 * netdev_WARN() acts like dev_printk(), but with the key difference 4809 * of using a WARN/WARN_ON to get the message out, including the 4810 * file/line information and a backtrace. 4811 */ 4812 #define netdev_WARN(dev, format, args...) \ 4813 WARN(1, "netdevice: %s%s: " format, netdev_name(dev), \ 4814 netdev_reg_state(dev), ##args) 4815 4816 #define netdev_WARN_ONCE(dev, format, args...) \ 4817 WARN_ONCE(1, "netdevice: %s%s: " format, netdev_name(dev), \ 4818 netdev_reg_state(dev), ##args) 4819 4820 /* netif printk helpers, similar to netdev_printk */ 4821 4822 #define netif_printk(priv, type, level, dev, fmt, args...) \ 4823 do { \ 4824 if (netif_msg_##type(priv)) \ 4825 netdev_printk(level, (dev), fmt, ##args); \ 4826 } while (0) 4827 4828 #define netif_level(level, priv, type, dev, fmt, args...) \ 4829 do { \ 4830 if (netif_msg_##type(priv)) \ 4831 netdev_##level(dev, fmt, ##args); \ 4832 } while (0) 4833 4834 #define netif_emerg(priv, type, dev, fmt, args...) \ 4835 netif_level(emerg, priv, type, dev, fmt, ##args) 4836 #define netif_alert(priv, type, dev, fmt, args...) \ 4837 netif_level(alert, priv, type, dev, fmt, ##args) 4838 #define netif_crit(priv, type, dev, fmt, args...) \ 4839 netif_level(crit, priv, type, dev, fmt, ##args) 4840 #define netif_err(priv, type, dev, fmt, args...) \ 4841 netif_level(err, priv, type, dev, fmt, ##args) 4842 #define netif_warn(priv, type, dev, fmt, args...) \ 4843 netif_level(warn, priv, type, dev, fmt, ##args) 4844 #define netif_notice(priv, type, dev, fmt, args...) \ 4845 netif_level(notice, priv, type, dev, fmt, ##args) 4846 #define netif_info(priv, type, dev, fmt, args...) \ 4847 netif_level(info, priv, type, dev, fmt, ##args) 4848 4849 #if defined(CONFIG_DYNAMIC_DEBUG) 4850 #define netif_dbg(priv, type, netdev, format, args...) \ 4851 do { \ 4852 if (netif_msg_##type(priv)) \ 4853 dynamic_netdev_dbg(netdev, format, ##args); \ 4854 } while (0) 4855 #elif defined(DEBUG) 4856 #define netif_dbg(priv, type, dev, format, args...) \ 4857 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args) 4858 #else 4859 #define netif_dbg(priv, type, dev, format, args...) \ 4860 ({ \ 4861 if (0) \ 4862 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \ 4863 0; \ 4864 }) 4865 #endif 4866 4867 /* if @cond then downgrade to debug, else print at @level */ 4868 #define netif_cond_dbg(priv, type, netdev, cond, level, fmt, args...) \ 4869 do { \ 4870 if (cond) \ 4871 netif_dbg(priv, type, netdev, fmt, ##args); \ 4872 else \ 4873 netif_ ## level(priv, type, netdev, fmt, ##args); \ 4874 } while (0) 4875 4876 #if defined(VERBOSE_DEBUG) 4877 #define netif_vdbg netif_dbg 4878 #else 4879 #define netif_vdbg(priv, type, dev, format, args...) \ 4880 ({ \ 4881 if (0) \ 4882 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \ 4883 0; \ 4884 }) 4885 #endif 4886 4887 /* 4888 * The list of packet types we will receive (as opposed to discard) 4889 * and the routines to invoke. 4890 * 4891 * Why 16. Because with 16 the only overlap we get on a hash of the 4892 * low nibble of the protocol value is RARP/SNAP/X.25. 4893 * 4894 * 0800 IP 4895 * 0001 802.3 4896 * 0002 AX.25 4897 * 0004 802.2 4898 * 8035 RARP 4899 * 0005 SNAP 4900 * 0805 X.25 4901 * 0806 ARP 4902 * 8137 IPX 4903 * 0009 Localtalk 4904 * 86DD IPv6 4905 */ 4906 #define PTYPE_HASH_SIZE (16) 4907 #define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1) 4908 4909 extern struct net_device *blackhole_netdev; 4910 4911 #endif /* _LINUX_NETDEVICE_H */ 4912