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, unsigned int txqueue); 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 unsigned int txqueue); 1286 1287 void (*ndo_get_stats64)(struct net_device *dev, 1288 struct rtnl_link_stats64 *storage); 1289 bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id); 1290 int (*ndo_get_offload_stats)(int attr_id, 1291 const struct net_device *dev, 1292 void *attr_data); 1293 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev); 1294 1295 int (*ndo_vlan_rx_add_vid)(struct net_device *dev, 1296 __be16 proto, u16 vid); 1297 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, 1298 __be16 proto, u16 vid); 1299 #ifdef CONFIG_NET_POLL_CONTROLLER 1300 void (*ndo_poll_controller)(struct net_device *dev); 1301 int (*ndo_netpoll_setup)(struct net_device *dev, 1302 struct netpoll_info *info); 1303 void (*ndo_netpoll_cleanup)(struct net_device *dev); 1304 #endif 1305 int (*ndo_set_vf_mac)(struct net_device *dev, 1306 int queue, u8 *mac); 1307 int (*ndo_set_vf_vlan)(struct net_device *dev, 1308 int queue, u16 vlan, 1309 u8 qos, __be16 proto); 1310 int (*ndo_set_vf_rate)(struct net_device *dev, 1311 int vf, int min_tx_rate, 1312 int max_tx_rate); 1313 int (*ndo_set_vf_spoofchk)(struct net_device *dev, 1314 int vf, bool setting); 1315 int (*ndo_set_vf_trust)(struct net_device *dev, 1316 int vf, bool setting); 1317 int (*ndo_get_vf_config)(struct net_device *dev, 1318 int vf, 1319 struct ifla_vf_info *ivf); 1320 int (*ndo_set_vf_link_state)(struct net_device *dev, 1321 int vf, int link_state); 1322 int (*ndo_get_vf_stats)(struct net_device *dev, 1323 int vf, 1324 struct ifla_vf_stats 1325 *vf_stats); 1326 int (*ndo_set_vf_port)(struct net_device *dev, 1327 int vf, 1328 struct nlattr *port[]); 1329 int (*ndo_get_vf_port)(struct net_device *dev, 1330 int vf, struct sk_buff *skb); 1331 int (*ndo_get_vf_guid)(struct net_device *dev, 1332 int vf, 1333 struct ifla_vf_guid *node_guid, 1334 struct ifla_vf_guid *port_guid); 1335 int (*ndo_set_vf_guid)(struct net_device *dev, 1336 int vf, u64 guid, 1337 int guid_type); 1338 int (*ndo_set_vf_rss_query_en)( 1339 struct net_device *dev, 1340 int vf, bool setting); 1341 int (*ndo_setup_tc)(struct net_device *dev, 1342 enum tc_setup_type type, 1343 void *type_data); 1344 #if IS_ENABLED(CONFIG_FCOE) 1345 int (*ndo_fcoe_enable)(struct net_device *dev); 1346 int (*ndo_fcoe_disable)(struct net_device *dev); 1347 int (*ndo_fcoe_ddp_setup)(struct net_device *dev, 1348 u16 xid, 1349 struct scatterlist *sgl, 1350 unsigned int sgc); 1351 int (*ndo_fcoe_ddp_done)(struct net_device *dev, 1352 u16 xid); 1353 int (*ndo_fcoe_ddp_target)(struct net_device *dev, 1354 u16 xid, 1355 struct scatterlist *sgl, 1356 unsigned int sgc); 1357 int (*ndo_fcoe_get_hbainfo)(struct net_device *dev, 1358 struct netdev_fcoe_hbainfo *hbainfo); 1359 #endif 1360 1361 #if IS_ENABLED(CONFIG_LIBFCOE) 1362 #define NETDEV_FCOE_WWNN 0 1363 #define NETDEV_FCOE_WWPN 1 1364 int (*ndo_fcoe_get_wwn)(struct net_device *dev, 1365 u64 *wwn, int type); 1366 #endif 1367 1368 #ifdef CONFIG_RFS_ACCEL 1369 int (*ndo_rx_flow_steer)(struct net_device *dev, 1370 const struct sk_buff *skb, 1371 u16 rxq_index, 1372 u32 flow_id); 1373 #endif 1374 int (*ndo_add_slave)(struct net_device *dev, 1375 struct net_device *slave_dev, 1376 struct netlink_ext_ack *extack); 1377 int (*ndo_del_slave)(struct net_device *dev, 1378 struct net_device *slave_dev); 1379 netdev_features_t (*ndo_fix_features)(struct net_device *dev, 1380 netdev_features_t features); 1381 int (*ndo_set_features)(struct net_device *dev, 1382 netdev_features_t features); 1383 int (*ndo_neigh_construct)(struct net_device *dev, 1384 struct neighbour *n); 1385 void (*ndo_neigh_destroy)(struct net_device *dev, 1386 struct neighbour *n); 1387 1388 int (*ndo_fdb_add)(struct ndmsg *ndm, 1389 struct nlattr *tb[], 1390 struct net_device *dev, 1391 const unsigned char *addr, 1392 u16 vid, 1393 u16 flags, 1394 struct netlink_ext_ack *extack); 1395 int (*ndo_fdb_del)(struct ndmsg *ndm, 1396 struct nlattr *tb[], 1397 struct net_device *dev, 1398 const unsigned char *addr, 1399 u16 vid); 1400 int (*ndo_fdb_dump)(struct sk_buff *skb, 1401 struct netlink_callback *cb, 1402 struct net_device *dev, 1403 struct net_device *filter_dev, 1404 int *idx); 1405 int (*ndo_fdb_get)(struct sk_buff *skb, 1406 struct nlattr *tb[], 1407 struct net_device *dev, 1408 const unsigned char *addr, 1409 u16 vid, u32 portid, u32 seq, 1410 struct netlink_ext_ack *extack); 1411 int (*ndo_bridge_setlink)(struct net_device *dev, 1412 struct nlmsghdr *nlh, 1413 u16 flags, 1414 struct netlink_ext_ack *extack); 1415 int (*ndo_bridge_getlink)(struct sk_buff *skb, 1416 u32 pid, u32 seq, 1417 struct net_device *dev, 1418 u32 filter_mask, 1419 int nlflags); 1420 int (*ndo_bridge_dellink)(struct net_device *dev, 1421 struct nlmsghdr *nlh, 1422 u16 flags); 1423 int (*ndo_change_carrier)(struct net_device *dev, 1424 bool new_carrier); 1425 int (*ndo_get_phys_port_id)(struct net_device *dev, 1426 struct netdev_phys_item_id *ppid); 1427 int (*ndo_get_port_parent_id)(struct net_device *dev, 1428 struct netdev_phys_item_id *ppid); 1429 int (*ndo_get_phys_port_name)(struct net_device *dev, 1430 char *name, size_t len); 1431 void (*ndo_udp_tunnel_add)(struct net_device *dev, 1432 struct udp_tunnel_info *ti); 1433 void (*ndo_udp_tunnel_del)(struct net_device *dev, 1434 struct udp_tunnel_info *ti); 1435 void* (*ndo_dfwd_add_station)(struct net_device *pdev, 1436 struct net_device *dev); 1437 void (*ndo_dfwd_del_station)(struct net_device *pdev, 1438 void *priv); 1439 1440 int (*ndo_set_tx_maxrate)(struct net_device *dev, 1441 int queue_index, 1442 u32 maxrate); 1443 int (*ndo_get_iflink)(const struct net_device *dev); 1444 int (*ndo_change_proto_down)(struct net_device *dev, 1445 bool proto_down); 1446 int (*ndo_fill_metadata_dst)(struct net_device *dev, 1447 struct sk_buff *skb); 1448 void (*ndo_set_rx_headroom)(struct net_device *dev, 1449 int needed_headroom); 1450 int (*ndo_bpf)(struct net_device *dev, 1451 struct netdev_bpf *bpf); 1452 int (*ndo_xdp_xmit)(struct net_device *dev, int n, 1453 struct xdp_frame **xdp, 1454 u32 flags); 1455 int (*ndo_xsk_wakeup)(struct net_device *dev, 1456 u32 queue_id, u32 flags); 1457 struct devlink_port * (*ndo_get_devlink_port)(struct net_device *dev); 1458 }; 1459 1460 /** 1461 * enum net_device_priv_flags - &struct net_device priv_flags 1462 * 1463 * These are the &struct net_device, they are only set internally 1464 * by drivers and used in the kernel. These flags are invisible to 1465 * userspace; this means that the order of these flags can change 1466 * during any kernel release. 1467 * 1468 * You should have a pretty good reason to be extending these flags. 1469 * 1470 * @IFF_802_1Q_VLAN: 802.1Q VLAN device 1471 * @IFF_EBRIDGE: Ethernet bridging device 1472 * @IFF_BONDING: bonding master or slave 1473 * @IFF_ISATAP: ISATAP interface (RFC4214) 1474 * @IFF_WAN_HDLC: WAN HDLC device 1475 * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to 1476 * release skb->dst 1477 * @IFF_DONT_BRIDGE: disallow bridging this ether dev 1478 * @IFF_DISABLE_NETPOLL: disable netpoll at run-time 1479 * @IFF_MACVLAN_PORT: device used as macvlan port 1480 * @IFF_BRIDGE_PORT: device used as bridge port 1481 * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port 1482 * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit 1483 * @IFF_UNICAST_FLT: Supports unicast filtering 1484 * @IFF_TEAM_PORT: device used as team port 1485 * @IFF_SUPP_NOFCS: device supports sending custom FCS 1486 * @IFF_LIVE_ADDR_CHANGE: device supports hardware address 1487 * change when it's running 1488 * @IFF_MACVLAN: Macvlan device 1489 * @IFF_XMIT_DST_RELEASE_PERM: IFF_XMIT_DST_RELEASE not taking into account 1490 * underlying stacked devices 1491 * @IFF_L3MDEV_MASTER: device is an L3 master device 1492 * @IFF_NO_QUEUE: device can run without qdisc attached 1493 * @IFF_OPENVSWITCH: device is a Open vSwitch master 1494 * @IFF_L3MDEV_SLAVE: device is enslaved to an L3 master device 1495 * @IFF_TEAM: device is a team device 1496 * @IFF_RXFH_CONFIGURED: device has had Rx Flow indirection table configured 1497 * @IFF_PHONY_HEADROOM: the headroom value is controlled by an external 1498 * entity (i.e. the master device for bridged veth) 1499 * @IFF_MACSEC: device is a MACsec device 1500 * @IFF_NO_RX_HANDLER: device doesn't support the rx_handler hook 1501 * @IFF_FAILOVER: device is a failover master device 1502 * @IFF_FAILOVER_SLAVE: device is lower dev of a failover master device 1503 * @IFF_L3MDEV_RX_HANDLER: only invoke the rx handler of L3 master device 1504 * @IFF_LIVE_RENAME_OK: rename is allowed while device is up and running 1505 */ 1506 enum netdev_priv_flags { 1507 IFF_802_1Q_VLAN = 1<<0, 1508 IFF_EBRIDGE = 1<<1, 1509 IFF_BONDING = 1<<2, 1510 IFF_ISATAP = 1<<3, 1511 IFF_WAN_HDLC = 1<<4, 1512 IFF_XMIT_DST_RELEASE = 1<<5, 1513 IFF_DONT_BRIDGE = 1<<6, 1514 IFF_DISABLE_NETPOLL = 1<<7, 1515 IFF_MACVLAN_PORT = 1<<8, 1516 IFF_BRIDGE_PORT = 1<<9, 1517 IFF_OVS_DATAPATH = 1<<10, 1518 IFF_TX_SKB_SHARING = 1<<11, 1519 IFF_UNICAST_FLT = 1<<12, 1520 IFF_TEAM_PORT = 1<<13, 1521 IFF_SUPP_NOFCS = 1<<14, 1522 IFF_LIVE_ADDR_CHANGE = 1<<15, 1523 IFF_MACVLAN = 1<<16, 1524 IFF_XMIT_DST_RELEASE_PERM = 1<<17, 1525 IFF_L3MDEV_MASTER = 1<<18, 1526 IFF_NO_QUEUE = 1<<19, 1527 IFF_OPENVSWITCH = 1<<20, 1528 IFF_L3MDEV_SLAVE = 1<<21, 1529 IFF_TEAM = 1<<22, 1530 IFF_RXFH_CONFIGURED = 1<<23, 1531 IFF_PHONY_HEADROOM = 1<<24, 1532 IFF_MACSEC = 1<<25, 1533 IFF_NO_RX_HANDLER = 1<<26, 1534 IFF_FAILOVER = 1<<27, 1535 IFF_FAILOVER_SLAVE = 1<<28, 1536 IFF_L3MDEV_RX_HANDLER = 1<<29, 1537 IFF_LIVE_RENAME_OK = 1<<30, 1538 }; 1539 1540 #define IFF_802_1Q_VLAN IFF_802_1Q_VLAN 1541 #define IFF_EBRIDGE IFF_EBRIDGE 1542 #define IFF_BONDING IFF_BONDING 1543 #define IFF_ISATAP IFF_ISATAP 1544 #define IFF_WAN_HDLC IFF_WAN_HDLC 1545 #define IFF_XMIT_DST_RELEASE IFF_XMIT_DST_RELEASE 1546 #define IFF_DONT_BRIDGE IFF_DONT_BRIDGE 1547 #define IFF_DISABLE_NETPOLL IFF_DISABLE_NETPOLL 1548 #define IFF_MACVLAN_PORT IFF_MACVLAN_PORT 1549 #define IFF_BRIDGE_PORT IFF_BRIDGE_PORT 1550 #define IFF_OVS_DATAPATH IFF_OVS_DATAPATH 1551 #define IFF_TX_SKB_SHARING IFF_TX_SKB_SHARING 1552 #define IFF_UNICAST_FLT IFF_UNICAST_FLT 1553 #define IFF_TEAM_PORT IFF_TEAM_PORT 1554 #define IFF_SUPP_NOFCS IFF_SUPP_NOFCS 1555 #define IFF_LIVE_ADDR_CHANGE IFF_LIVE_ADDR_CHANGE 1556 #define IFF_MACVLAN IFF_MACVLAN 1557 #define IFF_XMIT_DST_RELEASE_PERM IFF_XMIT_DST_RELEASE_PERM 1558 #define IFF_L3MDEV_MASTER IFF_L3MDEV_MASTER 1559 #define IFF_NO_QUEUE IFF_NO_QUEUE 1560 #define IFF_OPENVSWITCH IFF_OPENVSWITCH 1561 #define IFF_L3MDEV_SLAVE IFF_L3MDEV_SLAVE 1562 #define IFF_TEAM IFF_TEAM 1563 #define IFF_RXFH_CONFIGURED IFF_RXFH_CONFIGURED 1564 #define IFF_MACSEC IFF_MACSEC 1565 #define IFF_NO_RX_HANDLER IFF_NO_RX_HANDLER 1566 #define IFF_FAILOVER IFF_FAILOVER 1567 #define IFF_FAILOVER_SLAVE IFF_FAILOVER_SLAVE 1568 #define IFF_L3MDEV_RX_HANDLER IFF_L3MDEV_RX_HANDLER 1569 #define IFF_LIVE_RENAME_OK IFF_LIVE_RENAME_OK 1570 1571 /** 1572 * struct net_device - The DEVICE structure. 1573 * 1574 * Actually, this whole structure is a big mistake. It mixes I/O 1575 * data with strictly "high-level" data, and it has to know about 1576 * almost every data structure used in the INET module. 1577 * 1578 * @name: This is the first field of the "visible" part of this structure 1579 * (i.e. as seen by users in the "Space.c" file). It is the name 1580 * of the interface. 1581 * 1582 * @name_node: Name hashlist node 1583 * @ifalias: SNMP alias 1584 * @mem_end: Shared memory end 1585 * @mem_start: Shared memory start 1586 * @base_addr: Device I/O address 1587 * @irq: Device IRQ number 1588 * 1589 * @state: Generic network queuing layer state, see netdev_state_t 1590 * @dev_list: The global list of network devices 1591 * @napi_list: List entry used for polling NAPI devices 1592 * @unreg_list: List entry when we are unregistering the 1593 * device; see the function unregister_netdev 1594 * @close_list: List entry used when we are closing the device 1595 * @ptype_all: Device-specific packet handlers for all protocols 1596 * @ptype_specific: Device-specific, protocol-specific packet handlers 1597 * 1598 * @adj_list: Directly linked devices, like slaves for bonding 1599 * @features: Currently active device features 1600 * @hw_features: User-changeable features 1601 * 1602 * @wanted_features: User-requested features 1603 * @vlan_features: Mask of features inheritable by VLAN devices 1604 * 1605 * @hw_enc_features: Mask of features inherited by encapsulating devices 1606 * This field indicates what encapsulation 1607 * offloads the hardware is capable of doing, 1608 * and drivers will need to set them appropriately. 1609 * 1610 * @mpls_features: Mask of features inheritable by MPLS 1611 * 1612 * @ifindex: interface index 1613 * @group: The group the device belongs to 1614 * 1615 * @stats: Statistics struct, which was left as a legacy, use 1616 * rtnl_link_stats64 instead 1617 * 1618 * @rx_dropped: Dropped packets by core network, 1619 * do not use this in drivers 1620 * @tx_dropped: Dropped packets by core network, 1621 * do not use this in drivers 1622 * @rx_nohandler: nohandler dropped packets by core network on 1623 * inactive devices, do not use this in drivers 1624 * @carrier_up_count: Number of times the carrier has been up 1625 * @carrier_down_count: Number of times the carrier has been down 1626 * 1627 * @wireless_handlers: List of functions to handle Wireless Extensions, 1628 * instead of ioctl, 1629 * see <net/iw_handler.h> for details. 1630 * @wireless_data: Instance data managed by the core of wireless extensions 1631 * 1632 * @netdev_ops: Includes several pointers to callbacks, 1633 * if one wants to override the ndo_*() functions 1634 * @ethtool_ops: Management operations 1635 * @ndisc_ops: Includes callbacks for different IPv6 neighbour 1636 * discovery handling. Necessary for e.g. 6LoWPAN. 1637 * @header_ops: Includes callbacks for creating,parsing,caching,etc 1638 * of Layer 2 headers. 1639 * 1640 * @flags: Interface flags (a la BSD) 1641 * @priv_flags: Like 'flags' but invisible to userspace, 1642 * see if.h for the definitions 1643 * @gflags: Global flags ( kept as legacy ) 1644 * @padded: How much padding added by alloc_netdev() 1645 * @operstate: RFC2863 operstate 1646 * @link_mode: Mapping policy to operstate 1647 * @if_port: Selectable AUI, TP, ... 1648 * @dma: DMA channel 1649 * @mtu: Interface MTU value 1650 * @min_mtu: Interface Minimum MTU value 1651 * @max_mtu: Interface Maximum MTU value 1652 * @type: Interface hardware type 1653 * @hard_header_len: Maximum hardware header length. 1654 * @min_header_len: Minimum hardware header length 1655 * 1656 * @needed_headroom: Extra headroom the hardware may need, but not in all 1657 * cases can this be guaranteed 1658 * @needed_tailroom: Extra tailroom the hardware may need, but not in all 1659 * cases can this be guaranteed. Some cases also use 1660 * LL_MAX_HEADER instead to allocate the skb 1661 * 1662 * interface address info: 1663 * 1664 * @perm_addr: Permanent hw address 1665 * @addr_assign_type: Hw address assignment type 1666 * @addr_len: Hardware address length 1667 * @upper_level: Maximum depth level of upper devices. 1668 * @lower_level: Maximum depth level of lower devices. 1669 * @neigh_priv_len: Used in neigh_alloc() 1670 * @dev_id: Used to differentiate devices that share 1671 * the same link layer address 1672 * @dev_port: Used to differentiate devices that share 1673 * the same function 1674 * @addr_list_lock: XXX: need comments on this one 1675 * @uc_promisc: Counter that indicates promiscuous mode 1676 * has been enabled due to the need to listen to 1677 * additional unicast addresses in a device that 1678 * does not implement ndo_set_rx_mode() 1679 * @uc: unicast mac addresses 1680 * @mc: multicast mac addresses 1681 * @dev_addrs: list of device hw addresses 1682 * @queues_kset: Group of all Kobjects in the Tx and RX queues 1683 * @promiscuity: Number of times the NIC is told to work in 1684 * promiscuous mode; if it becomes 0 the NIC will 1685 * exit promiscuous mode 1686 * @allmulti: Counter, enables or disables allmulticast mode 1687 * 1688 * @vlan_info: VLAN info 1689 * @dsa_ptr: dsa specific data 1690 * @tipc_ptr: TIPC specific data 1691 * @atalk_ptr: AppleTalk link 1692 * @ip_ptr: IPv4 specific data 1693 * @dn_ptr: DECnet specific data 1694 * @ip6_ptr: IPv6 specific data 1695 * @ax25_ptr: AX.25 specific data 1696 * @ieee80211_ptr: IEEE 802.11 specific data, assign before registering 1697 * 1698 * @dev_addr: Hw address (before bcast, 1699 * because most packets are unicast) 1700 * 1701 * @_rx: Array of RX queues 1702 * @num_rx_queues: Number of RX queues 1703 * allocated at register_netdev() time 1704 * @real_num_rx_queues: Number of RX queues currently active in device 1705 * 1706 * @rx_handler: handler for received packets 1707 * @rx_handler_data: XXX: need comments on this one 1708 * @miniq_ingress: ingress/clsact qdisc specific data for 1709 * ingress processing 1710 * @ingress_queue: XXX: need comments on this one 1711 * @broadcast: hw bcast address 1712 * 1713 * @rx_cpu_rmap: CPU reverse-mapping for RX completion interrupts, 1714 * indexed by RX queue number. Assigned by driver. 1715 * This must only be set if the ndo_rx_flow_steer 1716 * operation is defined 1717 * @index_hlist: Device index hash chain 1718 * 1719 * @_tx: Array of TX queues 1720 * @num_tx_queues: Number of TX queues allocated at alloc_netdev_mq() time 1721 * @real_num_tx_queues: Number of TX queues currently active in device 1722 * @qdisc: Root qdisc from userspace point of view 1723 * @tx_queue_len: Max frames per queue allowed 1724 * @tx_global_lock: XXX: need comments on this one 1725 * 1726 * @xps_maps: XXX: need comments on this one 1727 * @miniq_egress: clsact qdisc specific data for 1728 * egress processing 1729 * @watchdog_timeo: Represents the timeout that is used by 1730 * the watchdog (see dev_watchdog()) 1731 * @watchdog_timer: List of timers 1732 * 1733 * @pcpu_refcnt: Number of references to this device 1734 * @todo_list: Delayed register/unregister 1735 * @link_watch_list: XXX: need comments on this one 1736 * 1737 * @reg_state: Register/unregister state machine 1738 * @dismantle: Device is going to be freed 1739 * @rtnl_link_state: This enum represents the phases of creating 1740 * a new link 1741 * 1742 * @needs_free_netdev: Should unregister perform free_netdev? 1743 * @priv_destructor: Called from unregister 1744 * @npinfo: XXX: need comments on this one 1745 * @nd_net: Network namespace this network device is inside 1746 * 1747 * @ml_priv: Mid-layer private 1748 * @lstats: Loopback statistics 1749 * @tstats: Tunnel statistics 1750 * @dstats: Dummy statistics 1751 * @vstats: Virtual ethernet statistics 1752 * 1753 * @garp_port: GARP 1754 * @mrp_port: MRP 1755 * 1756 * @dev: Class/net/name entry 1757 * @sysfs_groups: Space for optional device, statistics and wireless 1758 * sysfs groups 1759 * 1760 * @sysfs_rx_queue_group: Space for optional per-rx queue attributes 1761 * @rtnl_link_ops: Rtnl_link_ops 1762 * 1763 * @gso_max_size: Maximum size of generic segmentation offload 1764 * @gso_max_segs: Maximum number of segments that can be passed to the 1765 * NIC for GSO 1766 * 1767 * @dcbnl_ops: Data Center Bridging netlink ops 1768 * @num_tc: Number of traffic classes in the net device 1769 * @tc_to_txq: XXX: need comments on this one 1770 * @prio_tc_map: XXX: need comments on this one 1771 * 1772 * @fcoe_ddp_xid: Max exchange id for FCoE LRO by ddp 1773 * 1774 * @priomap: XXX: need comments on this one 1775 * @phydev: Physical device may attach itself 1776 * for hardware timestamping 1777 * @sfp_bus: attached &struct sfp_bus structure. 1778 * @qdisc_tx_busylock_key: lockdep class annotating Qdisc->busylock 1779 spinlock 1780 * @qdisc_running_key: lockdep class annotating Qdisc->running seqcount 1781 * @qdisc_xmit_lock_key: lockdep class annotating 1782 * netdev_queue->_xmit_lock spinlock 1783 * @addr_list_lock_key: lockdep class annotating 1784 * net_device->addr_list_lock spinlock 1785 * 1786 * @proto_down: protocol port state information can be sent to the 1787 * switch driver and used to set the phys state of the 1788 * switch port. 1789 * 1790 * @wol_enabled: Wake-on-LAN is enabled 1791 * 1792 * FIXME: cleanup struct net_device such that network protocol info 1793 * moves out. 1794 */ 1795 1796 struct net_device { 1797 char name[IFNAMSIZ]; 1798 struct netdev_name_node *name_node; 1799 struct dev_ifalias __rcu *ifalias; 1800 /* 1801 * I/O specific fields 1802 * FIXME: Merge these and struct ifmap into one 1803 */ 1804 unsigned long mem_end; 1805 unsigned long mem_start; 1806 unsigned long base_addr; 1807 int irq; 1808 1809 /* 1810 * Some hardware also needs these fields (state,dev_list, 1811 * napi_list,unreg_list,close_list) but they are not 1812 * part of the usual set specified in Space.c. 1813 */ 1814 1815 unsigned long state; 1816 1817 struct list_head dev_list; 1818 struct list_head napi_list; 1819 struct list_head unreg_list; 1820 struct list_head close_list; 1821 struct list_head ptype_all; 1822 struct list_head ptype_specific; 1823 1824 struct { 1825 struct list_head upper; 1826 struct list_head lower; 1827 } adj_list; 1828 1829 netdev_features_t features; 1830 netdev_features_t hw_features; 1831 netdev_features_t wanted_features; 1832 netdev_features_t vlan_features; 1833 netdev_features_t hw_enc_features; 1834 netdev_features_t mpls_features; 1835 netdev_features_t gso_partial_features; 1836 1837 int ifindex; 1838 int group; 1839 1840 struct net_device_stats stats; 1841 1842 atomic_long_t rx_dropped; 1843 atomic_long_t tx_dropped; 1844 atomic_long_t rx_nohandler; 1845 1846 /* Stats to monitor link on/off, flapping */ 1847 atomic_t carrier_up_count; 1848 atomic_t carrier_down_count; 1849 1850 #ifdef CONFIG_WIRELESS_EXT 1851 const struct iw_handler_def *wireless_handlers; 1852 struct iw_public_data *wireless_data; 1853 #endif 1854 const struct net_device_ops *netdev_ops; 1855 const struct ethtool_ops *ethtool_ops; 1856 #ifdef CONFIG_NET_L3_MASTER_DEV 1857 const struct l3mdev_ops *l3mdev_ops; 1858 #endif 1859 #if IS_ENABLED(CONFIG_IPV6) 1860 const struct ndisc_ops *ndisc_ops; 1861 #endif 1862 1863 #ifdef CONFIG_XFRM_OFFLOAD 1864 const struct xfrmdev_ops *xfrmdev_ops; 1865 #endif 1866 1867 #if IS_ENABLED(CONFIG_TLS_DEVICE) 1868 const struct tlsdev_ops *tlsdev_ops; 1869 #endif 1870 1871 const struct header_ops *header_ops; 1872 1873 unsigned int flags; 1874 unsigned int priv_flags; 1875 1876 unsigned short gflags; 1877 unsigned short padded; 1878 1879 unsigned char operstate; 1880 unsigned char link_mode; 1881 1882 unsigned char if_port; 1883 unsigned char dma; 1884 1885 /* Note : dev->mtu is often read without holding a lock. 1886 * Writers usually hold RTNL. 1887 * It is recommended to use READ_ONCE() to annotate the reads, 1888 * and to use WRITE_ONCE() to annotate the writes. 1889 */ 1890 unsigned int mtu; 1891 unsigned int min_mtu; 1892 unsigned int max_mtu; 1893 unsigned short type; 1894 unsigned short hard_header_len; 1895 unsigned char min_header_len; 1896 1897 unsigned short needed_headroom; 1898 unsigned short needed_tailroom; 1899 1900 /* Interface address info. */ 1901 unsigned char perm_addr[MAX_ADDR_LEN]; 1902 unsigned char addr_assign_type; 1903 unsigned char addr_len; 1904 unsigned char upper_level; 1905 unsigned char lower_level; 1906 unsigned short neigh_priv_len; 1907 unsigned short dev_id; 1908 unsigned short dev_port; 1909 spinlock_t addr_list_lock; 1910 unsigned char name_assign_type; 1911 bool uc_promisc; 1912 struct netdev_hw_addr_list uc; 1913 struct netdev_hw_addr_list mc; 1914 struct netdev_hw_addr_list dev_addrs; 1915 1916 #ifdef CONFIG_SYSFS 1917 struct kset *queues_kset; 1918 #endif 1919 unsigned int promiscuity; 1920 unsigned int allmulti; 1921 1922 1923 /* Protocol-specific pointers */ 1924 1925 #if IS_ENABLED(CONFIG_VLAN_8021Q) 1926 struct vlan_info __rcu *vlan_info; 1927 #endif 1928 #if IS_ENABLED(CONFIG_NET_DSA) 1929 struct dsa_port *dsa_ptr; 1930 #endif 1931 #if IS_ENABLED(CONFIG_TIPC) 1932 struct tipc_bearer __rcu *tipc_ptr; 1933 #endif 1934 #if IS_ENABLED(CONFIG_IRDA) || IS_ENABLED(CONFIG_ATALK) 1935 void *atalk_ptr; 1936 #endif 1937 struct in_device __rcu *ip_ptr; 1938 #if IS_ENABLED(CONFIG_DECNET) 1939 struct dn_dev __rcu *dn_ptr; 1940 #endif 1941 struct inet6_dev __rcu *ip6_ptr; 1942 #if IS_ENABLED(CONFIG_AX25) 1943 void *ax25_ptr; 1944 #endif 1945 struct wireless_dev *ieee80211_ptr; 1946 struct wpan_dev *ieee802154_ptr; 1947 #if IS_ENABLED(CONFIG_MPLS_ROUTING) 1948 struct mpls_dev __rcu *mpls_ptr; 1949 #endif 1950 1951 /* 1952 * Cache lines mostly used on receive path (including eth_type_trans()) 1953 */ 1954 /* Interface address info used in eth_type_trans() */ 1955 unsigned char *dev_addr; 1956 1957 struct netdev_rx_queue *_rx; 1958 unsigned int num_rx_queues; 1959 unsigned int real_num_rx_queues; 1960 1961 struct bpf_prog __rcu *xdp_prog; 1962 unsigned long gro_flush_timeout; 1963 rx_handler_func_t __rcu *rx_handler; 1964 void __rcu *rx_handler_data; 1965 1966 #ifdef CONFIG_NET_CLS_ACT 1967 struct mini_Qdisc __rcu *miniq_ingress; 1968 #endif 1969 struct netdev_queue __rcu *ingress_queue; 1970 #ifdef CONFIG_NETFILTER_INGRESS 1971 struct nf_hook_entries __rcu *nf_hooks_ingress; 1972 #endif 1973 1974 unsigned char broadcast[MAX_ADDR_LEN]; 1975 #ifdef CONFIG_RFS_ACCEL 1976 struct cpu_rmap *rx_cpu_rmap; 1977 #endif 1978 struct hlist_node index_hlist; 1979 1980 /* 1981 * Cache lines mostly used on transmit path 1982 */ 1983 struct netdev_queue *_tx ____cacheline_aligned_in_smp; 1984 unsigned int num_tx_queues; 1985 unsigned int real_num_tx_queues; 1986 struct Qdisc *qdisc; 1987 #ifdef CONFIG_NET_SCHED 1988 DECLARE_HASHTABLE (qdisc_hash, 4); 1989 #endif 1990 unsigned int tx_queue_len; 1991 spinlock_t tx_global_lock; 1992 int watchdog_timeo; 1993 1994 #ifdef CONFIG_XPS 1995 struct xps_dev_maps __rcu *xps_cpus_map; 1996 struct xps_dev_maps __rcu *xps_rxqs_map; 1997 #endif 1998 #ifdef CONFIG_NET_CLS_ACT 1999 struct mini_Qdisc __rcu *miniq_egress; 2000 #endif 2001 2002 /* These may be needed for future network-power-down code. */ 2003 struct timer_list watchdog_timer; 2004 2005 int __percpu *pcpu_refcnt; 2006 struct list_head todo_list; 2007 2008 struct list_head link_watch_list; 2009 2010 enum { NETREG_UNINITIALIZED=0, 2011 NETREG_REGISTERED, /* completed register_netdevice */ 2012 NETREG_UNREGISTERING, /* called unregister_netdevice */ 2013 NETREG_UNREGISTERED, /* completed unregister todo */ 2014 NETREG_RELEASED, /* called free_netdev */ 2015 NETREG_DUMMY, /* dummy device for NAPI poll */ 2016 } reg_state:8; 2017 2018 bool dismantle; 2019 2020 enum { 2021 RTNL_LINK_INITIALIZED, 2022 RTNL_LINK_INITIALIZING, 2023 } rtnl_link_state:16; 2024 2025 bool needs_free_netdev; 2026 void (*priv_destructor)(struct net_device *dev); 2027 2028 #ifdef CONFIG_NETPOLL 2029 struct netpoll_info __rcu *npinfo; 2030 #endif 2031 2032 possible_net_t nd_net; 2033 2034 /* mid-layer private */ 2035 union { 2036 void *ml_priv; 2037 struct pcpu_lstats __percpu *lstats; 2038 struct pcpu_sw_netstats __percpu *tstats; 2039 struct pcpu_dstats __percpu *dstats; 2040 }; 2041 2042 #if IS_ENABLED(CONFIG_GARP) 2043 struct garp_port __rcu *garp_port; 2044 #endif 2045 #if IS_ENABLED(CONFIG_MRP) 2046 struct mrp_port __rcu *mrp_port; 2047 #endif 2048 2049 struct device dev; 2050 const struct attribute_group *sysfs_groups[4]; 2051 const struct attribute_group *sysfs_rx_queue_group; 2052 2053 const struct rtnl_link_ops *rtnl_link_ops; 2054 2055 /* for setting kernel sock attribute on TCP connection setup */ 2056 #define GSO_MAX_SIZE 65536 2057 unsigned int gso_max_size; 2058 #define GSO_MAX_SEGS 65535 2059 u16 gso_max_segs; 2060 2061 #ifdef CONFIG_DCB 2062 const struct dcbnl_rtnl_ops *dcbnl_ops; 2063 #endif 2064 s16 num_tc; 2065 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE]; 2066 u8 prio_tc_map[TC_BITMASK + 1]; 2067 2068 #if IS_ENABLED(CONFIG_FCOE) 2069 unsigned int fcoe_ddp_xid; 2070 #endif 2071 #if IS_ENABLED(CONFIG_CGROUP_NET_PRIO) 2072 struct netprio_map __rcu *priomap; 2073 #endif 2074 struct phy_device *phydev; 2075 struct sfp_bus *sfp_bus; 2076 struct lock_class_key qdisc_tx_busylock_key; 2077 struct lock_class_key qdisc_running_key; 2078 struct lock_class_key qdisc_xmit_lock_key; 2079 struct lock_class_key addr_list_lock_key; 2080 bool proto_down; 2081 unsigned wol_enabled:1; 2082 }; 2083 #define to_net_dev(d) container_of(d, struct net_device, dev) 2084 2085 static inline bool netif_elide_gro(const struct net_device *dev) 2086 { 2087 if (!(dev->features & NETIF_F_GRO) || dev->xdp_prog) 2088 return true; 2089 return false; 2090 } 2091 2092 #define NETDEV_ALIGN 32 2093 2094 static inline 2095 int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio) 2096 { 2097 return dev->prio_tc_map[prio & TC_BITMASK]; 2098 } 2099 2100 static inline 2101 int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc) 2102 { 2103 if (tc >= dev->num_tc) 2104 return -EINVAL; 2105 2106 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK; 2107 return 0; 2108 } 2109 2110 int netdev_txq_to_tc(struct net_device *dev, unsigned int txq); 2111 void netdev_reset_tc(struct net_device *dev); 2112 int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset); 2113 int netdev_set_num_tc(struct net_device *dev, u8 num_tc); 2114 2115 static inline 2116 int netdev_get_num_tc(struct net_device *dev) 2117 { 2118 return dev->num_tc; 2119 } 2120 2121 void netdev_unbind_sb_channel(struct net_device *dev, 2122 struct net_device *sb_dev); 2123 int netdev_bind_sb_channel_queue(struct net_device *dev, 2124 struct net_device *sb_dev, 2125 u8 tc, u16 count, u16 offset); 2126 int netdev_set_sb_channel(struct net_device *dev, u16 channel); 2127 static inline int netdev_get_sb_channel(struct net_device *dev) 2128 { 2129 return max_t(int, -dev->num_tc, 0); 2130 } 2131 2132 static inline 2133 struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev, 2134 unsigned int index) 2135 { 2136 return &dev->_tx[index]; 2137 } 2138 2139 static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev, 2140 const struct sk_buff *skb) 2141 { 2142 return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb)); 2143 } 2144 2145 static inline void netdev_for_each_tx_queue(struct net_device *dev, 2146 void (*f)(struct net_device *, 2147 struct netdev_queue *, 2148 void *), 2149 void *arg) 2150 { 2151 unsigned int i; 2152 2153 for (i = 0; i < dev->num_tx_queues; i++) 2154 f(dev, &dev->_tx[i], arg); 2155 } 2156 2157 u16 netdev_pick_tx(struct net_device *dev, struct sk_buff *skb, 2158 struct net_device *sb_dev); 2159 struct netdev_queue *netdev_core_pick_tx(struct net_device *dev, 2160 struct sk_buff *skb, 2161 struct net_device *sb_dev); 2162 2163 /* returns the headroom that the master device needs to take in account 2164 * when forwarding to this dev 2165 */ 2166 static inline unsigned netdev_get_fwd_headroom(struct net_device *dev) 2167 { 2168 return dev->priv_flags & IFF_PHONY_HEADROOM ? 0 : dev->needed_headroom; 2169 } 2170 2171 static inline void netdev_set_rx_headroom(struct net_device *dev, int new_hr) 2172 { 2173 if (dev->netdev_ops->ndo_set_rx_headroom) 2174 dev->netdev_ops->ndo_set_rx_headroom(dev, new_hr); 2175 } 2176 2177 /* set the device rx headroom to the dev's default */ 2178 static inline void netdev_reset_rx_headroom(struct net_device *dev) 2179 { 2180 netdev_set_rx_headroom(dev, -1); 2181 } 2182 2183 /* 2184 * Net namespace inlines 2185 */ 2186 static inline 2187 struct net *dev_net(const struct net_device *dev) 2188 { 2189 return read_pnet(&dev->nd_net); 2190 } 2191 2192 static inline 2193 void dev_net_set(struct net_device *dev, struct net *net) 2194 { 2195 write_pnet(&dev->nd_net, net); 2196 } 2197 2198 /** 2199 * netdev_priv - access network device private data 2200 * @dev: network device 2201 * 2202 * Get network device private data 2203 */ 2204 static inline void *netdev_priv(const struct net_device *dev) 2205 { 2206 return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN); 2207 } 2208 2209 /* Set the sysfs physical device reference for the network logical device 2210 * if set prior to registration will cause a symlink during initialization. 2211 */ 2212 #define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev)) 2213 2214 /* Set the sysfs device type for the network logical device to allow 2215 * fine-grained identification of different network device types. For 2216 * example Ethernet, Wireless LAN, Bluetooth, WiMAX etc. 2217 */ 2218 #define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype)) 2219 2220 /* Default NAPI poll() weight 2221 * Device drivers are strongly advised to not use bigger value 2222 */ 2223 #define NAPI_POLL_WEIGHT 64 2224 2225 /** 2226 * netif_napi_add - initialize a NAPI context 2227 * @dev: network device 2228 * @napi: NAPI context 2229 * @poll: polling function 2230 * @weight: default weight 2231 * 2232 * netif_napi_add() must be used to initialize a NAPI context prior to calling 2233 * *any* of the other NAPI-related functions. 2234 */ 2235 void netif_napi_add(struct net_device *dev, struct napi_struct *napi, 2236 int (*poll)(struct napi_struct *, int), int weight); 2237 2238 /** 2239 * netif_tx_napi_add - initialize a NAPI context 2240 * @dev: network device 2241 * @napi: NAPI context 2242 * @poll: polling function 2243 * @weight: default weight 2244 * 2245 * This variant of netif_napi_add() should be used from drivers using NAPI 2246 * to exclusively poll a TX queue. 2247 * This will avoid we add it into napi_hash[], thus polluting this hash table. 2248 */ 2249 static inline void netif_tx_napi_add(struct net_device *dev, 2250 struct napi_struct *napi, 2251 int (*poll)(struct napi_struct *, int), 2252 int weight) 2253 { 2254 set_bit(NAPI_STATE_NO_BUSY_POLL, &napi->state); 2255 netif_napi_add(dev, napi, poll, weight); 2256 } 2257 2258 /** 2259 * netif_napi_del - remove a NAPI context 2260 * @napi: NAPI context 2261 * 2262 * netif_napi_del() removes a NAPI context from the network device NAPI list 2263 */ 2264 void netif_napi_del(struct napi_struct *napi); 2265 2266 struct napi_gro_cb { 2267 /* Virtual address of skb_shinfo(skb)->frags[0].page + offset. */ 2268 void *frag0; 2269 2270 /* Length of frag0. */ 2271 unsigned int frag0_len; 2272 2273 /* This indicates where we are processing relative to skb->data. */ 2274 int data_offset; 2275 2276 /* This is non-zero if the packet cannot be merged with the new skb. */ 2277 u16 flush; 2278 2279 /* Save the IP ID here and check when we get to the transport layer */ 2280 u16 flush_id; 2281 2282 /* Number of segments aggregated. */ 2283 u16 count; 2284 2285 /* Start offset for remote checksum offload */ 2286 u16 gro_remcsum_start; 2287 2288 /* jiffies when first packet was created/queued */ 2289 unsigned long age; 2290 2291 /* Used in ipv6_gro_receive() and foo-over-udp */ 2292 u16 proto; 2293 2294 /* This is non-zero if the packet may be of the same flow. */ 2295 u8 same_flow:1; 2296 2297 /* Used in tunnel GRO receive */ 2298 u8 encap_mark:1; 2299 2300 /* GRO checksum is valid */ 2301 u8 csum_valid:1; 2302 2303 /* Number of checksums via CHECKSUM_UNNECESSARY */ 2304 u8 csum_cnt:3; 2305 2306 /* Free the skb? */ 2307 u8 free:2; 2308 #define NAPI_GRO_FREE 1 2309 #define NAPI_GRO_FREE_STOLEN_HEAD 2 2310 2311 /* Used in foo-over-udp, set in udp[46]_gro_receive */ 2312 u8 is_ipv6:1; 2313 2314 /* Used in GRE, set in fou/gue_gro_receive */ 2315 u8 is_fou:1; 2316 2317 /* Used to determine if flush_id can be ignored */ 2318 u8 is_atomic:1; 2319 2320 /* Number of gro_receive callbacks this packet already went through */ 2321 u8 recursion_counter:4; 2322 2323 /* 1 bit hole */ 2324 2325 /* used to support CHECKSUM_COMPLETE for tunneling protocols */ 2326 __wsum csum; 2327 2328 /* used in skb_gro_receive() slow path */ 2329 struct sk_buff *last; 2330 }; 2331 2332 #define NAPI_GRO_CB(skb) ((struct napi_gro_cb *)(skb)->cb) 2333 2334 #define GRO_RECURSION_LIMIT 15 2335 static inline int gro_recursion_inc_test(struct sk_buff *skb) 2336 { 2337 return ++NAPI_GRO_CB(skb)->recursion_counter == GRO_RECURSION_LIMIT; 2338 } 2339 2340 typedef struct sk_buff *(*gro_receive_t)(struct list_head *, struct sk_buff *); 2341 static inline struct sk_buff *call_gro_receive(gro_receive_t cb, 2342 struct list_head *head, 2343 struct sk_buff *skb) 2344 { 2345 if (unlikely(gro_recursion_inc_test(skb))) { 2346 NAPI_GRO_CB(skb)->flush |= 1; 2347 return NULL; 2348 } 2349 2350 return cb(head, skb); 2351 } 2352 2353 typedef struct sk_buff *(*gro_receive_sk_t)(struct sock *, struct list_head *, 2354 struct sk_buff *); 2355 static inline struct sk_buff *call_gro_receive_sk(gro_receive_sk_t cb, 2356 struct sock *sk, 2357 struct list_head *head, 2358 struct sk_buff *skb) 2359 { 2360 if (unlikely(gro_recursion_inc_test(skb))) { 2361 NAPI_GRO_CB(skb)->flush |= 1; 2362 return NULL; 2363 } 2364 2365 return cb(sk, head, skb); 2366 } 2367 2368 struct packet_type { 2369 __be16 type; /* This is really htons(ether_type). */ 2370 bool ignore_outgoing; 2371 struct net_device *dev; /* NULL is wildcarded here */ 2372 int (*func) (struct sk_buff *, 2373 struct net_device *, 2374 struct packet_type *, 2375 struct net_device *); 2376 void (*list_func) (struct list_head *, 2377 struct packet_type *, 2378 struct net_device *); 2379 bool (*id_match)(struct packet_type *ptype, 2380 struct sock *sk); 2381 void *af_packet_priv; 2382 struct list_head list; 2383 }; 2384 2385 struct offload_callbacks { 2386 struct sk_buff *(*gso_segment)(struct sk_buff *skb, 2387 netdev_features_t features); 2388 struct sk_buff *(*gro_receive)(struct list_head *head, 2389 struct sk_buff *skb); 2390 int (*gro_complete)(struct sk_buff *skb, int nhoff); 2391 }; 2392 2393 struct packet_offload { 2394 __be16 type; /* This is really htons(ether_type). */ 2395 u16 priority; 2396 struct offload_callbacks callbacks; 2397 struct list_head list; 2398 }; 2399 2400 /* often modified stats are per-CPU, other are shared (netdev->stats) */ 2401 struct pcpu_sw_netstats { 2402 u64 rx_packets; 2403 u64 rx_bytes; 2404 u64 tx_packets; 2405 u64 tx_bytes; 2406 struct u64_stats_sync syncp; 2407 } __aligned(4 * sizeof(u64)); 2408 2409 struct pcpu_lstats { 2410 u64_stats_t packets; 2411 u64_stats_t bytes; 2412 struct u64_stats_sync syncp; 2413 } __aligned(2 * sizeof(u64)); 2414 2415 void dev_lstats_read(struct net_device *dev, u64 *packets, u64 *bytes); 2416 2417 static inline void dev_lstats_add(struct net_device *dev, unsigned int len) 2418 { 2419 struct pcpu_lstats *lstats = this_cpu_ptr(dev->lstats); 2420 2421 u64_stats_update_begin(&lstats->syncp); 2422 u64_stats_add(&lstats->bytes, len); 2423 u64_stats_inc(&lstats->packets); 2424 u64_stats_update_end(&lstats->syncp); 2425 } 2426 2427 #define __netdev_alloc_pcpu_stats(type, gfp) \ 2428 ({ \ 2429 typeof(type) __percpu *pcpu_stats = alloc_percpu_gfp(type, gfp);\ 2430 if (pcpu_stats) { \ 2431 int __cpu; \ 2432 for_each_possible_cpu(__cpu) { \ 2433 typeof(type) *stat; \ 2434 stat = per_cpu_ptr(pcpu_stats, __cpu); \ 2435 u64_stats_init(&stat->syncp); \ 2436 } \ 2437 } \ 2438 pcpu_stats; \ 2439 }) 2440 2441 #define netdev_alloc_pcpu_stats(type) \ 2442 __netdev_alloc_pcpu_stats(type, GFP_KERNEL) 2443 2444 enum netdev_lag_tx_type { 2445 NETDEV_LAG_TX_TYPE_UNKNOWN, 2446 NETDEV_LAG_TX_TYPE_RANDOM, 2447 NETDEV_LAG_TX_TYPE_BROADCAST, 2448 NETDEV_LAG_TX_TYPE_ROUNDROBIN, 2449 NETDEV_LAG_TX_TYPE_ACTIVEBACKUP, 2450 NETDEV_LAG_TX_TYPE_HASH, 2451 }; 2452 2453 enum netdev_lag_hash { 2454 NETDEV_LAG_HASH_NONE, 2455 NETDEV_LAG_HASH_L2, 2456 NETDEV_LAG_HASH_L34, 2457 NETDEV_LAG_HASH_L23, 2458 NETDEV_LAG_HASH_E23, 2459 NETDEV_LAG_HASH_E34, 2460 NETDEV_LAG_HASH_UNKNOWN, 2461 }; 2462 2463 struct netdev_lag_upper_info { 2464 enum netdev_lag_tx_type tx_type; 2465 enum netdev_lag_hash hash_type; 2466 }; 2467 2468 struct netdev_lag_lower_state_info { 2469 u8 link_up : 1, 2470 tx_enabled : 1; 2471 }; 2472 2473 #include <linux/notifier.h> 2474 2475 /* netdevice notifier chain. Please remember to update netdev_cmd_to_name() 2476 * and the rtnetlink notification exclusion list in rtnetlink_event() when 2477 * adding new types. 2478 */ 2479 enum netdev_cmd { 2480 NETDEV_UP = 1, /* For now you can't veto a device up/down */ 2481 NETDEV_DOWN, 2482 NETDEV_REBOOT, /* Tell a protocol stack a network interface 2483 detected a hardware crash and restarted 2484 - we can use this eg to kick tcp sessions 2485 once done */ 2486 NETDEV_CHANGE, /* Notify device state change */ 2487 NETDEV_REGISTER, 2488 NETDEV_UNREGISTER, 2489 NETDEV_CHANGEMTU, /* notify after mtu change happened */ 2490 NETDEV_CHANGEADDR, /* notify after the address change */ 2491 NETDEV_PRE_CHANGEADDR, /* notify before the address change */ 2492 NETDEV_GOING_DOWN, 2493 NETDEV_CHANGENAME, 2494 NETDEV_FEAT_CHANGE, 2495 NETDEV_BONDING_FAILOVER, 2496 NETDEV_PRE_UP, 2497 NETDEV_PRE_TYPE_CHANGE, 2498 NETDEV_POST_TYPE_CHANGE, 2499 NETDEV_POST_INIT, 2500 NETDEV_RELEASE, 2501 NETDEV_NOTIFY_PEERS, 2502 NETDEV_JOIN, 2503 NETDEV_CHANGEUPPER, 2504 NETDEV_RESEND_IGMP, 2505 NETDEV_PRECHANGEMTU, /* notify before mtu change happened */ 2506 NETDEV_CHANGEINFODATA, 2507 NETDEV_BONDING_INFO, 2508 NETDEV_PRECHANGEUPPER, 2509 NETDEV_CHANGELOWERSTATE, 2510 NETDEV_UDP_TUNNEL_PUSH_INFO, 2511 NETDEV_UDP_TUNNEL_DROP_INFO, 2512 NETDEV_CHANGE_TX_QUEUE_LEN, 2513 NETDEV_CVLAN_FILTER_PUSH_INFO, 2514 NETDEV_CVLAN_FILTER_DROP_INFO, 2515 NETDEV_SVLAN_FILTER_PUSH_INFO, 2516 NETDEV_SVLAN_FILTER_DROP_INFO, 2517 }; 2518 const char *netdev_cmd_to_name(enum netdev_cmd cmd); 2519 2520 int register_netdevice_notifier(struct notifier_block *nb); 2521 int unregister_netdevice_notifier(struct notifier_block *nb); 2522 int register_netdevice_notifier_net(struct net *net, struct notifier_block *nb); 2523 int unregister_netdevice_notifier_net(struct net *net, 2524 struct notifier_block *nb); 2525 2526 struct netdev_notifier_info { 2527 struct net_device *dev; 2528 struct netlink_ext_ack *extack; 2529 }; 2530 2531 struct netdev_notifier_info_ext { 2532 struct netdev_notifier_info info; /* must be first */ 2533 union { 2534 u32 mtu; 2535 } ext; 2536 }; 2537 2538 struct netdev_notifier_change_info { 2539 struct netdev_notifier_info info; /* must be first */ 2540 unsigned int flags_changed; 2541 }; 2542 2543 struct netdev_notifier_changeupper_info { 2544 struct netdev_notifier_info info; /* must be first */ 2545 struct net_device *upper_dev; /* new upper dev */ 2546 bool master; /* is upper dev master */ 2547 bool linking; /* is the notification for link or unlink */ 2548 void *upper_info; /* upper dev info */ 2549 }; 2550 2551 struct netdev_notifier_changelowerstate_info { 2552 struct netdev_notifier_info info; /* must be first */ 2553 void *lower_state_info; /* is lower dev state */ 2554 }; 2555 2556 struct netdev_notifier_pre_changeaddr_info { 2557 struct netdev_notifier_info info; /* must be first */ 2558 const unsigned char *dev_addr; 2559 }; 2560 2561 static inline void netdev_notifier_info_init(struct netdev_notifier_info *info, 2562 struct net_device *dev) 2563 { 2564 info->dev = dev; 2565 info->extack = NULL; 2566 } 2567 2568 static inline struct net_device * 2569 netdev_notifier_info_to_dev(const struct netdev_notifier_info *info) 2570 { 2571 return info->dev; 2572 } 2573 2574 static inline struct netlink_ext_ack * 2575 netdev_notifier_info_to_extack(const struct netdev_notifier_info *info) 2576 { 2577 return info->extack; 2578 } 2579 2580 int call_netdevice_notifiers(unsigned long val, struct net_device *dev); 2581 2582 2583 extern rwlock_t dev_base_lock; /* Device list lock */ 2584 2585 #define for_each_netdev(net, d) \ 2586 list_for_each_entry(d, &(net)->dev_base_head, dev_list) 2587 #define for_each_netdev_reverse(net, d) \ 2588 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list) 2589 #define for_each_netdev_rcu(net, d) \ 2590 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list) 2591 #define for_each_netdev_safe(net, d, n) \ 2592 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list) 2593 #define for_each_netdev_continue(net, d) \ 2594 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list) 2595 #define for_each_netdev_continue_reverse(net, d) \ 2596 list_for_each_entry_continue_reverse(d, &(net)->dev_base_head, \ 2597 dev_list) 2598 #define for_each_netdev_continue_rcu(net, d) \ 2599 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list) 2600 #define for_each_netdev_in_bond_rcu(bond, slave) \ 2601 for_each_netdev_rcu(&init_net, slave) \ 2602 if (netdev_master_upper_dev_get_rcu(slave) == (bond)) 2603 #define net_device_entry(lh) list_entry(lh, struct net_device, dev_list) 2604 2605 static inline struct net_device *next_net_device(struct net_device *dev) 2606 { 2607 struct list_head *lh; 2608 struct net *net; 2609 2610 net = dev_net(dev); 2611 lh = dev->dev_list.next; 2612 return lh == &net->dev_base_head ? NULL : net_device_entry(lh); 2613 } 2614 2615 static inline struct net_device *next_net_device_rcu(struct net_device *dev) 2616 { 2617 struct list_head *lh; 2618 struct net *net; 2619 2620 net = dev_net(dev); 2621 lh = rcu_dereference(list_next_rcu(&dev->dev_list)); 2622 return lh == &net->dev_base_head ? NULL : net_device_entry(lh); 2623 } 2624 2625 static inline struct net_device *first_net_device(struct net *net) 2626 { 2627 return list_empty(&net->dev_base_head) ? NULL : 2628 net_device_entry(net->dev_base_head.next); 2629 } 2630 2631 static inline struct net_device *first_net_device_rcu(struct net *net) 2632 { 2633 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head)); 2634 2635 return lh == &net->dev_base_head ? NULL : net_device_entry(lh); 2636 } 2637 2638 int netdev_boot_setup_check(struct net_device *dev); 2639 unsigned long netdev_boot_base(const char *prefix, int unit); 2640 struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type, 2641 const char *hwaddr); 2642 struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type); 2643 struct net_device *__dev_getfirstbyhwtype(struct net *net, unsigned short type); 2644 void dev_add_pack(struct packet_type *pt); 2645 void dev_remove_pack(struct packet_type *pt); 2646 void __dev_remove_pack(struct packet_type *pt); 2647 void dev_add_offload(struct packet_offload *po); 2648 void dev_remove_offload(struct packet_offload *po); 2649 2650 int dev_get_iflink(const struct net_device *dev); 2651 int dev_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb); 2652 struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags, 2653 unsigned short mask); 2654 struct net_device *dev_get_by_name(struct net *net, const char *name); 2655 struct net_device *dev_get_by_name_rcu(struct net *net, const char *name); 2656 struct net_device *__dev_get_by_name(struct net *net, const char *name); 2657 int dev_alloc_name(struct net_device *dev, const char *name); 2658 int dev_open(struct net_device *dev, struct netlink_ext_ack *extack); 2659 void dev_close(struct net_device *dev); 2660 void dev_close_many(struct list_head *head, bool unlink); 2661 void dev_disable_lro(struct net_device *dev); 2662 int dev_loopback_xmit(struct net *net, struct sock *sk, struct sk_buff *newskb); 2663 u16 dev_pick_tx_zero(struct net_device *dev, struct sk_buff *skb, 2664 struct net_device *sb_dev); 2665 u16 dev_pick_tx_cpu_id(struct net_device *dev, struct sk_buff *skb, 2666 struct net_device *sb_dev); 2667 int dev_queue_xmit(struct sk_buff *skb); 2668 int dev_queue_xmit_accel(struct sk_buff *skb, struct net_device *sb_dev); 2669 int dev_direct_xmit(struct sk_buff *skb, u16 queue_id); 2670 int register_netdevice(struct net_device *dev); 2671 void unregister_netdevice_queue(struct net_device *dev, struct list_head *head); 2672 void unregister_netdevice_many(struct list_head *head); 2673 static inline void unregister_netdevice(struct net_device *dev) 2674 { 2675 unregister_netdevice_queue(dev, NULL); 2676 } 2677 2678 int netdev_refcnt_read(const struct net_device *dev); 2679 void free_netdev(struct net_device *dev); 2680 void netdev_freemem(struct net_device *dev); 2681 void synchronize_net(void); 2682 int init_dummy_netdev(struct net_device *dev); 2683 2684 struct net_device *dev_get_by_index(struct net *net, int ifindex); 2685 struct net_device *__dev_get_by_index(struct net *net, int ifindex); 2686 struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex); 2687 struct net_device *dev_get_by_napi_id(unsigned int napi_id); 2688 int netdev_get_name(struct net *net, char *name, int ifindex); 2689 int dev_restart(struct net_device *dev); 2690 int skb_gro_receive(struct sk_buff *p, struct sk_buff *skb); 2691 2692 static inline unsigned int skb_gro_offset(const struct sk_buff *skb) 2693 { 2694 return NAPI_GRO_CB(skb)->data_offset; 2695 } 2696 2697 static inline unsigned int skb_gro_len(const struct sk_buff *skb) 2698 { 2699 return skb->len - NAPI_GRO_CB(skb)->data_offset; 2700 } 2701 2702 static inline void skb_gro_pull(struct sk_buff *skb, unsigned int len) 2703 { 2704 NAPI_GRO_CB(skb)->data_offset += len; 2705 } 2706 2707 static inline void *skb_gro_header_fast(struct sk_buff *skb, 2708 unsigned int offset) 2709 { 2710 return NAPI_GRO_CB(skb)->frag0 + offset; 2711 } 2712 2713 static inline int skb_gro_header_hard(struct sk_buff *skb, unsigned int hlen) 2714 { 2715 return NAPI_GRO_CB(skb)->frag0_len < hlen; 2716 } 2717 2718 static inline void skb_gro_frag0_invalidate(struct sk_buff *skb) 2719 { 2720 NAPI_GRO_CB(skb)->frag0 = NULL; 2721 NAPI_GRO_CB(skb)->frag0_len = 0; 2722 } 2723 2724 static inline void *skb_gro_header_slow(struct sk_buff *skb, unsigned int hlen, 2725 unsigned int offset) 2726 { 2727 if (!pskb_may_pull(skb, hlen)) 2728 return NULL; 2729 2730 skb_gro_frag0_invalidate(skb); 2731 return skb->data + offset; 2732 } 2733 2734 static inline void *skb_gro_network_header(struct sk_buff *skb) 2735 { 2736 return (NAPI_GRO_CB(skb)->frag0 ?: skb->data) + 2737 skb_network_offset(skb); 2738 } 2739 2740 static inline void skb_gro_postpull_rcsum(struct sk_buff *skb, 2741 const void *start, unsigned int len) 2742 { 2743 if (NAPI_GRO_CB(skb)->csum_valid) 2744 NAPI_GRO_CB(skb)->csum = csum_sub(NAPI_GRO_CB(skb)->csum, 2745 csum_partial(start, len, 0)); 2746 } 2747 2748 /* GRO checksum functions. These are logical equivalents of the normal 2749 * checksum functions (in skbuff.h) except that they operate on the GRO 2750 * offsets and fields in sk_buff. 2751 */ 2752 2753 __sum16 __skb_gro_checksum_complete(struct sk_buff *skb); 2754 2755 static inline bool skb_at_gro_remcsum_start(struct sk_buff *skb) 2756 { 2757 return (NAPI_GRO_CB(skb)->gro_remcsum_start == skb_gro_offset(skb)); 2758 } 2759 2760 static inline bool __skb_gro_checksum_validate_needed(struct sk_buff *skb, 2761 bool zero_okay, 2762 __sum16 check) 2763 { 2764 return ((skb->ip_summed != CHECKSUM_PARTIAL || 2765 skb_checksum_start_offset(skb) < 2766 skb_gro_offset(skb)) && 2767 !skb_at_gro_remcsum_start(skb) && 2768 NAPI_GRO_CB(skb)->csum_cnt == 0 && 2769 (!zero_okay || check)); 2770 } 2771 2772 static inline __sum16 __skb_gro_checksum_validate_complete(struct sk_buff *skb, 2773 __wsum psum) 2774 { 2775 if (NAPI_GRO_CB(skb)->csum_valid && 2776 !csum_fold(csum_add(psum, NAPI_GRO_CB(skb)->csum))) 2777 return 0; 2778 2779 NAPI_GRO_CB(skb)->csum = psum; 2780 2781 return __skb_gro_checksum_complete(skb); 2782 } 2783 2784 static inline void skb_gro_incr_csum_unnecessary(struct sk_buff *skb) 2785 { 2786 if (NAPI_GRO_CB(skb)->csum_cnt > 0) { 2787 /* Consume a checksum from CHECKSUM_UNNECESSARY */ 2788 NAPI_GRO_CB(skb)->csum_cnt--; 2789 } else { 2790 /* Update skb for CHECKSUM_UNNECESSARY and csum_level when we 2791 * verified a new top level checksum or an encapsulated one 2792 * during GRO. This saves work if we fallback to normal path. 2793 */ 2794 __skb_incr_checksum_unnecessary(skb); 2795 } 2796 } 2797 2798 #define __skb_gro_checksum_validate(skb, proto, zero_okay, check, \ 2799 compute_pseudo) \ 2800 ({ \ 2801 __sum16 __ret = 0; \ 2802 if (__skb_gro_checksum_validate_needed(skb, zero_okay, check)) \ 2803 __ret = __skb_gro_checksum_validate_complete(skb, \ 2804 compute_pseudo(skb, proto)); \ 2805 if (!__ret) \ 2806 skb_gro_incr_csum_unnecessary(skb); \ 2807 __ret; \ 2808 }) 2809 2810 #define skb_gro_checksum_validate(skb, proto, compute_pseudo) \ 2811 __skb_gro_checksum_validate(skb, proto, false, 0, compute_pseudo) 2812 2813 #define skb_gro_checksum_validate_zero_check(skb, proto, check, \ 2814 compute_pseudo) \ 2815 __skb_gro_checksum_validate(skb, proto, true, check, compute_pseudo) 2816 2817 #define skb_gro_checksum_simple_validate(skb) \ 2818 __skb_gro_checksum_validate(skb, 0, false, 0, null_compute_pseudo) 2819 2820 static inline bool __skb_gro_checksum_convert_check(struct sk_buff *skb) 2821 { 2822 return (NAPI_GRO_CB(skb)->csum_cnt == 0 && 2823 !NAPI_GRO_CB(skb)->csum_valid); 2824 } 2825 2826 static inline void __skb_gro_checksum_convert(struct sk_buff *skb, 2827 __sum16 check, __wsum pseudo) 2828 { 2829 NAPI_GRO_CB(skb)->csum = ~pseudo; 2830 NAPI_GRO_CB(skb)->csum_valid = 1; 2831 } 2832 2833 #define skb_gro_checksum_try_convert(skb, proto, check, compute_pseudo) \ 2834 do { \ 2835 if (__skb_gro_checksum_convert_check(skb)) \ 2836 __skb_gro_checksum_convert(skb, check, \ 2837 compute_pseudo(skb, proto)); \ 2838 } while (0) 2839 2840 struct gro_remcsum { 2841 int offset; 2842 __wsum delta; 2843 }; 2844 2845 static inline void skb_gro_remcsum_init(struct gro_remcsum *grc) 2846 { 2847 grc->offset = 0; 2848 grc->delta = 0; 2849 } 2850 2851 static inline void *skb_gro_remcsum_process(struct sk_buff *skb, void *ptr, 2852 unsigned int off, size_t hdrlen, 2853 int start, int offset, 2854 struct gro_remcsum *grc, 2855 bool nopartial) 2856 { 2857 __wsum delta; 2858 size_t plen = hdrlen + max_t(size_t, offset + sizeof(u16), start); 2859 2860 BUG_ON(!NAPI_GRO_CB(skb)->csum_valid); 2861 2862 if (!nopartial) { 2863 NAPI_GRO_CB(skb)->gro_remcsum_start = off + hdrlen + start; 2864 return ptr; 2865 } 2866 2867 ptr = skb_gro_header_fast(skb, off); 2868 if (skb_gro_header_hard(skb, off + plen)) { 2869 ptr = skb_gro_header_slow(skb, off + plen, off); 2870 if (!ptr) 2871 return NULL; 2872 } 2873 2874 delta = remcsum_adjust(ptr + hdrlen, NAPI_GRO_CB(skb)->csum, 2875 start, offset); 2876 2877 /* Adjust skb->csum since we changed the packet */ 2878 NAPI_GRO_CB(skb)->csum = csum_add(NAPI_GRO_CB(skb)->csum, delta); 2879 2880 grc->offset = off + hdrlen + offset; 2881 grc->delta = delta; 2882 2883 return ptr; 2884 } 2885 2886 static inline void skb_gro_remcsum_cleanup(struct sk_buff *skb, 2887 struct gro_remcsum *grc) 2888 { 2889 void *ptr; 2890 size_t plen = grc->offset + sizeof(u16); 2891 2892 if (!grc->delta) 2893 return; 2894 2895 ptr = skb_gro_header_fast(skb, grc->offset); 2896 if (skb_gro_header_hard(skb, grc->offset + sizeof(u16))) { 2897 ptr = skb_gro_header_slow(skb, plen, grc->offset); 2898 if (!ptr) 2899 return; 2900 } 2901 2902 remcsum_unadjust((__sum16 *)ptr, grc->delta); 2903 } 2904 2905 #ifdef CONFIG_XFRM_OFFLOAD 2906 static inline void skb_gro_flush_final(struct sk_buff *skb, struct sk_buff *pp, int flush) 2907 { 2908 if (PTR_ERR(pp) != -EINPROGRESS) 2909 NAPI_GRO_CB(skb)->flush |= flush; 2910 } 2911 static inline void skb_gro_flush_final_remcsum(struct sk_buff *skb, 2912 struct sk_buff *pp, 2913 int flush, 2914 struct gro_remcsum *grc) 2915 { 2916 if (PTR_ERR(pp) != -EINPROGRESS) { 2917 NAPI_GRO_CB(skb)->flush |= flush; 2918 skb_gro_remcsum_cleanup(skb, grc); 2919 skb->remcsum_offload = 0; 2920 } 2921 } 2922 #else 2923 static inline void skb_gro_flush_final(struct sk_buff *skb, struct sk_buff *pp, int flush) 2924 { 2925 NAPI_GRO_CB(skb)->flush |= flush; 2926 } 2927 static inline void skb_gro_flush_final_remcsum(struct sk_buff *skb, 2928 struct sk_buff *pp, 2929 int flush, 2930 struct gro_remcsum *grc) 2931 { 2932 NAPI_GRO_CB(skb)->flush |= flush; 2933 skb_gro_remcsum_cleanup(skb, grc); 2934 skb->remcsum_offload = 0; 2935 } 2936 #endif 2937 2938 static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev, 2939 unsigned short type, 2940 const void *daddr, const void *saddr, 2941 unsigned int len) 2942 { 2943 if (!dev->header_ops || !dev->header_ops->create) 2944 return 0; 2945 2946 return dev->header_ops->create(skb, dev, type, daddr, saddr, len); 2947 } 2948 2949 static inline int dev_parse_header(const struct sk_buff *skb, 2950 unsigned char *haddr) 2951 { 2952 const struct net_device *dev = skb->dev; 2953 2954 if (!dev->header_ops || !dev->header_ops->parse) 2955 return 0; 2956 return dev->header_ops->parse(skb, haddr); 2957 } 2958 2959 static inline __be16 dev_parse_header_protocol(const struct sk_buff *skb) 2960 { 2961 const struct net_device *dev = skb->dev; 2962 2963 if (!dev->header_ops || !dev->header_ops->parse_protocol) 2964 return 0; 2965 return dev->header_ops->parse_protocol(skb); 2966 } 2967 2968 /* ll_header must have at least hard_header_len allocated */ 2969 static inline bool dev_validate_header(const struct net_device *dev, 2970 char *ll_header, int len) 2971 { 2972 if (likely(len >= dev->hard_header_len)) 2973 return true; 2974 if (len < dev->min_header_len) 2975 return false; 2976 2977 if (capable(CAP_SYS_RAWIO)) { 2978 memset(ll_header + len, 0, dev->hard_header_len - len); 2979 return true; 2980 } 2981 2982 if (dev->header_ops && dev->header_ops->validate) 2983 return dev->header_ops->validate(ll_header, len); 2984 2985 return false; 2986 } 2987 2988 typedef int gifconf_func_t(struct net_device * dev, char __user * bufptr, 2989 int len, int size); 2990 int register_gifconf(unsigned int family, gifconf_func_t *gifconf); 2991 static inline int unregister_gifconf(unsigned int family) 2992 { 2993 return register_gifconf(family, NULL); 2994 } 2995 2996 #ifdef CONFIG_NET_FLOW_LIMIT 2997 #define FLOW_LIMIT_HISTORY (1 << 7) /* must be ^2 and !overflow buckets */ 2998 struct sd_flow_limit { 2999 u64 count; 3000 unsigned int num_buckets; 3001 unsigned int history_head; 3002 u16 history[FLOW_LIMIT_HISTORY]; 3003 u8 buckets[]; 3004 }; 3005 3006 extern int netdev_flow_limit_table_len; 3007 #endif /* CONFIG_NET_FLOW_LIMIT */ 3008 3009 /* 3010 * Incoming packets are placed on per-CPU queues 3011 */ 3012 struct softnet_data { 3013 struct list_head poll_list; 3014 struct sk_buff_head process_queue; 3015 3016 /* stats */ 3017 unsigned int processed; 3018 unsigned int time_squeeze; 3019 unsigned int received_rps; 3020 #ifdef CONFIG_RPS 3021 struct softnet_data *rps_ipi_list; 3022 #endif 3023 #ifdef CONFIG_NET_FLOW_LIMIT 3024 struct sd_flow_limit __rcu *flow_limit; 3025 #endif 3026 struct Qdisc *output_queue; 3027 struct Qdisc **output_queue_tailp; 3028 struct sk_buff *completion_queue; 3029 #ifdef CONFIG_XFRM_OFFLOAD 3030 struct sk_buff_head xfrm_backlog; 3031 #endif 3032 /* written and read only by owning cpu: */ 3033 struct { 3034 u16 recursion; 3035 u8 more; 3036 } xmit; 3037 #ifdef CONFIG_RPS 3038 /* input_queue_head should be written by cpu owning this struct, 3039 * and only read by other cpus. Worth using a cache line. 3040 */ 3041 unsigned int input_queue_head ____cacheline_aligned_in_smp; 3042 3043 /* Elements below can be accessed between CPUs for RPS/RFS */ 3044 call_single_data_t csd ____cacheline_aligned_in_smp; 3045 struct softnet_data *rps_ipi_next; 3046 unsigned int cpu; 3047 unsigned int input_queue_tail; 3048 #endif 3049 unsigned int dropped; 3050 struct sk_buff_head input_pkt_queue; 3051 struct napi_struct backlog; 3052 3053 }; 3054 3055 static inline void input_queue_head_incr(struct softnet_data *sd) 3056 { 3057 #ifdef CONFIG_RPS 3058 sd->input_queue_head++; 3059 #endif 3060 } 3061 3062 static inline void input_queue_tail_incr_save(struct softnet_data *sd, 3063 unsigned int *qtail) 3064 { 3065 #ifdef CONFIG_RPS 3066 *qtail = ++sd->input_queue_tail; 3067 #endif 3068 } 3069 3070 DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data); 3071 3072 static inline int dev_recursion_level(void) 3073 { 3074 return this_cpu_read(softnet_data.xmit.recursion); 3075 } 3076 3077 #define XMIT_RECURSION_LIMIT 10 3078 static inline bool dev_xmit_recursion(void) 3079 { 3080 return unlikely(__this_cpu_read(softnet_data.xmit.recursion) > 3081 XMIT_RECURSION_LIMIT); 3082 } 3083 3084 static inline void dev_xmit_recursion_inc(void) 3085 { 3086 __this_cpu_inc(softnet_data.xmit.recursion); 3087 } 3088 3089 static inline void dev_xmit_recursion_dec(void) 3090 { 3091 __this_cpu_dec(softnet_data.xmit.recursion); 3092 } 3093 3094 void __netif_schedule(struct Qdisc *q); 3095 void netif_schedule_queue(struct netdev_queue *txq); 3096 3097 static inline void netif_tx_schedule_all(struct net_device *dev) 3098 { 3099 unsigned int i; 3100 3101 for (i = 0; i < dev->num_tx_queues; i++) 3102 netif_schedule_queue(netdev_get_tx_queue(dev, i)); 3103 } 3104 3105 static __always_inline void netif_tx_start_queue(struct netdev_queue *dev_queue) 3106 { 3107 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state); 3108 } 3109 3110 /** 3111 * netif_start_queue - allow transmit 3112 * @dev: network device 3113 * 3114 * Allow upper layers to call the device hard_start_xmit routine. 3115 */ 3116 static inline void netif_start_queue(struct net_device *dev) 3117 { 3118 netif_tx_start_queue(netdev_get_tx_queue(dev, 0)); 3119 } 3120 3121 static inline void netif_tx_start_all_queues(struct net_device *dev) 3122 { 3123 unsigned int i; 3124 3125 for (i = 0; i < dev->num_tx_queues; i++) { 3126 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 3127 netif_tx_start_queue(txq); 3128 } 3129 } 3130 3131 void netif_tx_wake_queue(struct netdev_queue *dev_queue); 3132 3133 /** 3134 * netif_wake_queue - restart transmit 3135 * @dev: network device 3136 * 3137 * Allow upper layers to call the device hard_start_xmit routine. 3138 * Used for flow control when transmit resources are available. 3139 */ 3140 static inline void netif_wake_queue(struct net_device *dev) 3141 { 3142 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0)); 3143 } 3144 3145 static inline void netif_tx_wake_all_queues(struct net_device *dev) 3146 { 3147 unsigned int i; 3148 3149 for (i = 0; i < dev->num_tx_queues; i++) { 3150 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 3151 netif_tx_wake_queue(txq); 3152 } 3153 } 3154 3155 static __always_inline void netif_tx_stop_queue(struct netdev_queue *dev_queue) 3156 { 3157 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state); 3158 } 3159 3160 /** 3161 * netif_stop_queue - stop transmitted packets 3162 * @dev: network device 3163 * 3164 * Stop upper layers calling the device hard_start_xmit routine. 3165 * Used for flow control when transmit resources are unavailable. 3166 */ 3167 static inline void netif_stop_queue(struct net_device *dev) 3168 { 3169 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0)); 3170 } 3171 3172 void netif_tx_stop_all_queues(struct net_device *dev); 3173 void netdev_update_lockdep_key(struct net_device *dev); 3174 3175 static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue) 3176 { 3177 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state); 3178 } 3179 3180 /** 3181 * netif_queue_stopped - test if transmit queue is flowblocked 3182 * @dev: network device 3183 * 3184 * Test if transmit queue on device is currently unable to send. 3185 */ 3186 static inline bool netif_queue_stopped(const struct net_device *dev) 3187 { 3188 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0)); 3189 } 3190 3191 static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue) 3192 { 3193 return dev_queue->state & QUEUE_STATE_ANY_XOFF; 3194 } 3195 3196 static inline bool 3197 netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue) 3198 { 3199 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN; 3200 } 3201 3202 static inline bool 3203 netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue) 3204 { 3205 return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN; 3206 } 3207 3208 /** 3209 * netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write 3210 * @dev_queue: pointer to transmit queue 3211 * 3212 * BQL enabled drivers might use this helper in their ndo_start_xmit(), 3213 * to give appropriate hint to the CPU. 3214 */ 3215 static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue) 3216 { 3217 #ifdef CONFIG_BQL 3218 prefetchw(&dev_queue->dql.num_queued); 3219 #endif 3220 } 3221 3222 /** 3223 * netdev_txq_bql_complete_prefetchw - prefetch bql data for write 3224 * @dev_queue: pointer to transmit queue 3225 * 3226 * BQL enabled drivers might use this helper in their TX completion path, 3227 * to give appropriate hint to the CPU. 3228 */ 3229 static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue) 3230 { 3231 #ifdef CONFIG_BQL 3232 prefetchw(&dev_queue->dql.limit); 3233 #endif 3234 } 3235 3236 static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue, 3237 unsigned int bytes) 3238 { 3239 #ifdef CONFIG_BQL 3240 dql_queued(&dev_queue->dql, bytes); 3241 3242 if (likely(dql_avail(&dev_queue->dql) >= 0)) 3243 return; 3244 3245 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state); 3246 3247 /* 3248 * The XOFF flag must be set before checking the dql_avail below, 3249 * because in netdev_tx_completed_queue we update the dql_completed 3250 * before checking the XOFF flag. 3251 */ 3252 smp_mb(); 3253 3254 /* check again in case another CPU has just made room avail */ 3255 if (unlikely(dql_avail(&dev_queue->dql) >= 0)) 3256 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state); 3257 #endif 3258 } 3259 3260 /* Variant of netdev_tx_sent_queue() for drivers that are aware 3261 * that they should not test BQL status themselves. 3262 * We do want to change __QUEUE_STATE_STACK_XOFF only for the last 3263 * skb of a batch. 3264 * Returns true if the doorbell must be used to kick the NIC. 3265 */ 3266 static inline bool __netdev_tx_sent_queue(struct netdev_queue *dev_queue, 3267 unsigned int bytes, 3268 bool xmit_more) 3269 { 3270 if (xmit_more) { 3271 #ifdef CONFIG_BQL 3272 dql_queued(&dev_queue->dql, bytes); 3273 #endif 3274 return netif_tx_queue_stopped(dev_queue); 3275 } 3276 netdev_tx_sent_queue(dev_queue, bytes); 3277 return true; 3278 } 3279 3280 /** 3281 * netdev_sent_queue - report the number of bytes queued to hardware 3282 * @dev: network device 3283 * @bytes: number of bytes queued to the hardware device queue 3284 * 3285 * Report the number of bytes queued for sending/completion to the network 3286 * device hardware queue. @bytes should be a good approximation and should 3287 * exactly match netdev_completed_queue() @bytes 3288 */ 3289 static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes) 3290 { 3291 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes); 3292 } 3293 3294 static inline bool __netdev_sent_queue(struct net_device *dev, 3295 unsigned int bytes, 3296 bool xmit_more) 3297 { 3298 return __netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes, 3299 xmit_more); 3300 } 3301 3302 static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue, 3303 unsigned int pkts, unsigned int bytes) 3304 { 3305 #ifdef CONFIG_BQL 3306 if (unlikely(!bytes)) 3307 return; 3308 3309 dql_completed(&dev_queue->dql, bytes); 3310 3311 /* 3312 * Without the memory barrier there is a small possiblity that 3313 * netdev_tx_sent_queue will miss the update and cause the queue to 3314 * be stopped forever 3315 */ 3316 smp_mb(); 3317 3318 if (unlikely(dql_avail(&dev_queue->dql) < 0)) 3319 return; 3320 3321 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state)) 3322 netif_schedule_queue(dev_queue); 3323 #endif 3324 } 3325 3326 /** 3327 * netdev_completed_queue - report bytes and packets completed by device 3328 * @dev: network device 3329 * @pkts: actual number of packets sent over the medium 3330 * @bytes: actual number of bytes sent over the medium 3331 * 3332 * Report the number of bytes and packets transmitted by the network device 3333 * hardware queue over the physical medium, @bytes must exactly match the 3334 * @bytes amount passed to netdev_sent_queue() 3335 */ 3336 static inline void netdev_completed_queue(struct net_device *dev, 3337 unsigned int pkts, unsigned int bytes) 3338 { 3339 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes); 3340 } 3341 3342 static inline void netdev_tx_reset_queue(struct netdev_queue *q) 3343 { 3344 #ifdef CONFIG_BQL 3345 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state); 3346 dql_reset(&q->dql); 3347 #endif 3348 } 3349 3350 /** 3351 * netdev_reset_queue - reset the packets and bytes count of a network device 3352 * @dev_queue: network device 3353 * 3354 * Reset the bytes and packet count of a network device and clear the 3355 * software flow control OFF bit for this network device 3356 */ 3357 static inline void netdev_reset_queue(struct net_device *dev_queue) 3358 { 3359 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0)); 3360 } 3361 3362 /** 3363 * netdev_cap_txqueue - check if selected tx queue exceeds device queues 3364 * @dev: network device 3365 * @queue_index: given tx queue index 3366 * 3367 * Returns 0 if given tx queue index >= number of device tx queues, 3368 * otherwise returns the originally passed tx queue index. 3369 */ 3370 static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index) 3371 { 3372 if (unlikely(queue_index >= dev->real_num_tx_queues)) { 3373 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n", 3374 dev->name, queue_index, 3375 dev->real_num_tx_queues); 3376 return 0; 3377 } 3378 3379 return queue_index; 3380 } 3381 3382 /** 3383 * netif_running - test if up 3384 * @dev: network device 3385 * 3386 * Test if the device has been brought up. 3387 */ 3388 static inline bool netif_running(const struct net_device *dev) 3389 { 3390 return test_bit(__LINK_STATE_START, &dev->state); 3391 } 3392 3393 /* 3394 * Routines to manage the subqueues on a device. We only need start, 3395 * stop, and a check if it's stopped. All other device management is 3396 * done at the overall netdevice level. 3397 * Also test the device if we're multiqueue. 3398 */ 3399 3400 /** 3401 * netif_start_subqueue - allow sending packets on subqueue 3402 * @dev: network device 3403 * @queue_index: sub queue index 3404 * 3405 * Start individual transmit queue of a device with multiple transmit queues. 3406 */ 3407 static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index) 3408 { 3409 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 3410 3411 netif_tx_start_queue(txq); 3412 } 3413 3414 /** 3415 * netif_stop_subqueue - stop sending packets on subqueue 3416 * @dev: network device 3417 * @queue_index: sub queue index 3418 * 3419 * Stop individual transmit queue of a device with multiple transmit queues. 3420 */ 3421 static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index) 3422 { 3423 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 3424 netif_tx_stop_queue(txq); 3425 } 3426 3427 /** 3428 * netif_subqueue_stopped - test status of subqueue 3429 * @dev: network device 3430 * @queue_index: sub queue index 3431 * 3432 * Check individual transmit queue of a device with multiple transmit queues. 3433 */ 3434 static inline bool __netif_subqueue_stopped(const struct net_device *dev, 3435 u16 queue_index) 3436 { 3437 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 3438 3439 return netif_tx_queue_stopped(txq); 3440 } 3441 3442 static inline bool netif_subqueue_stopped(const struct net_device *dev, 3443 struct sk_buff *skb) 3444 { 3445 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb)); 3446 } 3447 3448 /** 3449 * netif_wake_subqueue - allow sending packets on subqueue 3450 * @dev: network device 3451 * @queue_index: sub queue index 3452 * 3453 * Resume individual transmit queue of a device with multiple transmit queues. 3454 */ 3455 static inline void netif_wake_subqueue(struct net_device *dev, u16 queue_index) 3456 { 3457 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 3458 3459 netif_tx_wake_queue(txq); 3460 } 3461 3462 #ifdef CONFIG_XPS 3463 int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask, 3464 u16 index); 3465 int __netif_set_xps_queue(struct net_device *dev, const unsigned long *mask, 3466 u16 index, bool is_rxqs_map); 3467 3468 /** 3469 * netif_attr_test_mask - Test a CPU or Rx queue set in a mask 3470 * @j: CPU/Rx queue index 3471 * @mask: bitmask of all cpus/rx queues 3472 * @nr_bits: number of bits in the bitmask 3473 * 3474 * Test if a CPU or Rx queue index is set in a mask of all CPU/Rx queues. 3475 */ 3476 static inline bool netif_attr_test_mask(unsigned long j, 3477 const unsigned long *mask, 3478 unsigned int nr_bits) 3479 { 3480 cpu_max_bits_warn(j, nr_bits); 3481 return test_bit(j, mask); 3482 } 3483 3484 /** 3485 * netif_attr_test_online - Test for online CPU/Rx queue 3486 * @j: CPU/Rx queue index 3487 * @online_mask: bitmask for CPUs/Rx queues that are online 3488 * @nr_bits: number of bits in the bitmask 3489 * 3490 * Returns true if a CPU/Rx queue is online. 3491 */ 3492 static inline bool netif_attr_test_online(unsigned long j, 3493 const unsigned long *online_mask, 3494 unsigned int nr_bits) 3495 { 3496 cpu_max_bits_warn(j, nr_bits); 3497 3498 if (online_mask) 3499 return test_bit(j, online_mask); 3500 3501 return (j < nr_bits); 3502 } 3503 3504 /** 3505 * netif_attrmask_next - get the next CPU/Rx queue in a cpu/Rx queues mask 3506 * @n: CPU/Rx queue index 3507 * @srcp: the cpumask/Rx queue mask pointer 3508 * @nr_bits: number of bits in the bitmask 3509 * 3510 * Returns >= nr_bits if no further CPUs/Rx queues set. 3511 */ 3512 static inline unsigned int netif_attrmask_next(int n, const unsigned long *srcp, 3513 unsigned int nr_bits) 3514 { 3515 /* -1 is a legal arg here. */ 3516 if (n != -1) 3517 cpu_max_bits_warn(n, nr_bits); 3518 3519 if (srcp) 3520 return find_next_bit(srcp, nr_bits, n + 1); 3521 3522 return n + 1; 3523 } 3524 3525 /** 3526 * netif_attrmask_next_and - get the next CPU/Rx queue in *src1p & *src2p 3527 * @n: CPU/Rx queue index 3528 * @src1p: the first CPUs/Rx queues mask pointer 3529 * @src2p: the second CPUs/Rx queues mask pointer 3530 * @nr_bits: number of bits in the bitmask 3531 * 3532 * Returns >= nr_bits if no further CPUs/Rx queues set in both. 3533 */ 3534 static inline int netif_attrmask_next_and(int n, const unsigned long *src1p, 3535 const unsigned long *src2p, 3536 unsigned int nr_bits) 3537 { 3538 /* -1 is a legal arg here. */ 3539 if (n != -1) 3540 cpu_max_bits_warn(n, nr_bits); 3541 3542 if (src1p && src2p) 3543 return find_next_and_bit(src1p, src2p, nr_bits, n + 1); 3544 else if (src1p) 3545 return find_next_bit(src1p, nr_bits, n + 1); 3546 else if (src2p) 3547 return find_next_bit(src2p, nr_bits, n + 1); 3548 3549 return n + 1; 3550 } 3551 #else 3552 static inline int netif_set_xps_queue(struct net_device *dev, 3553 const struct cpumask *mask, 3554 u16 index) 3555 { 3556 return 0; 3557 } 3558 3559 static inline int __netif_set_xps_queue(struct net_device *dev, 3560 const unsigned long *mask, 3561 u16 index, bool is_rxqs_map) 3562 { 3563 return 0; 3564 } 3565 #endif 3566 3567 /** 3568 * netif_is_multiqueue - test if device has multiple transmit queues 3569 * @dev: network device 3570 * 3571 * Check if device has multiple transmit queues 3572 */ 3573 static inline bool netif_is_multiqueue(const struct net_device *dev) 3574 { 3575 return dev->num_tx_queues > 1; 3576 } 3577 3578 int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq); 3579 3580 #ifdef CONFIG_SYSFS 3581 int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq); 3582 #else 3583 static inline int netif_set_real_num_rx_queues(struct net_device *dev, 3584 unsigned int rxqs) 3585 { 3586 dev->real_num_rx_queues = rxqs; 3587 return 0; 3588 } 3589 #endif 3590 3591 static inline struct netdev_rx_queue * 3592 __netif_get_rx_queue(struct net_device *dev, unsigned int rxq) 3593 { 3594 return dev->_rx + rxq; 3595 } 3596 3597 #ifdef CONFIG_SYSFS 3598 static inline unsigned int get_netdev_rx_queue_index( 3599 struct netdev_rx_queue *queue) 3600 { 3601 struct net_device *dev = queue->dev; 3602 int index = queue - dev->_rx; 3603 3604 BUG_ON(index >= dev->num_rx_queues); 3605 return index; 3606 } 3607 #endif 3608 3609 #define DEFAULT_MAX_NUM_RSS_QUEUES (8) 3610 int netif_get_num_default_rss_queues(void); 3611 3612 enum skb_free_reason { 3613 SKB_REASON_CONSUMED, 3614 SKB_REASON_DROPPED, 3615 }; 3616 3617 void __dev_kfree_skb_irq(struct sk_buff *skb, enum skb_free_reason reason); 3618 void __dev_kfree_skb_any(struct sk_buff *skb, enum skb_free_reason reason); 3619 3620 /* 3621 * It is not allowed to call kfree_skb() or consume_skb() from hardware 3622 * interrupt context or with hardware interrupts being disabled. 3623 * (in_irq() || irqs_disabled()) 3624 * 3625 * We provide four helpers that can be used in following contexts : 3626 * 3627 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context, 3628 * replacing kfree_skb(skb) 3629 * 3630 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context. 3631 * Typically used in place of consume_skb(skb) in TX completion path 3632 * 3633 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context, 3634 * replacing kfree_skb(skb) 3635 * 3636 * dev_consume_skb_any(skb) when caller doesn't know its current irq context, 3637 * and consumed a packet. Used in place of consume_skb(skb) 3638 */ 3639 static inline void dev_kfree_skb_irq(struct sk_buff *skb) 3640 { 3641 __dev_kfree_skb_irq(skb, SKB_REASON_DROPPED); 3642 } 3643 3644 static inline void dev_consume_skb_irq(struct sk_buff *skb) 3645 { 3646 __dev_kfree_skb_irq(skb, SKB_REASON_CONSUMED); 3647 } 3648 3649 static inline void dev_kfree_skb_any(struct sk_buff *skb) 3650 { 3651 __dev_kfree_skb_any(skb, SKB_REASON_DROPPED); 3652 } 3653 3654 static inline void dev_consume_skb_any(struct sk_buff *skb) 3655 { 3656 __dev_kfree_skb_any(skb, SKB_REASON_CONSUMED); 3657 } 3658 3659 void generic_xdp_tx(struct sk_buff *skb, struct bpf_prog *xdp_prog); 3660 int do_xdp_generic(struct bpf_prog *xdp_prog, struct sk_buff *skb); 3661 int netif_rx(struct sk_buff *skb); 3662 int netif_rx_ni(struct sk_buff *skb); 3663 int netif_receive_skb(struct sk_buff *skb); 3664 int netif_receive_skb_core(struct sk_buff *skb); 3665 void netif_receive_skb_list(struct list_head *head); 3666 gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb); 3667 void napi_gro_flush(struct napi_struct *napi, bool flush_old); 3668 struct sk_buff *napi_get_frags(struct napi_struct *napi); 3669 gro_result_t napi_gro_frags(struct napi_struct *napi); 3670 struct packet_offload *gro_find_receive_by_type(__be16 type); 3671 struct packet_offload *gro_find_complete_by_type(__be16 type); 3672 3673 static inline void napi_free_frags(struct napi_struct *napi) 3674 { 3675 kfree_skb(napi->skb); 3676 napi->skb = NULL; 3677 } 3678 3679 bool netdev_is_rx_handler_busy(struct net_device *dev); 3680 int netdev_rx_handler_register(struct net_device *dev, 3681 rx_handler_func_t *rx_handler, 3682 void *rx_handler_data); 3683 void netdev_rx_handler_unregister(struct net_device *dev); 3684 3685 bool dev_valid_name(const char *name); 3686 int dev_ioctl(struct net *net, unsigned int cmd, struct ifreq *ifr, 3687 bool *need_copyout); 3688 int dev_ifconf(struct net *net, struct ifconf *, int); 3689 int dev_ethtool(struct net *net, struct ifreq *); 3690 unsigned int dev_get_flags(const struct net_device *); 3691 int __dev_change_flags(struct net_device *dev, unsigned int flags, 3692 struct netlink_ext_ack *extack); 3693 int dev_change_flags(struct net_device *dev, unsigned int flags, 3694 struct netlink_ext_ack *extack); 3695 void __dev_notify_flags(struct net_device *, unsigned int old_flags, 3696 unsigned int gchanges); 3697 int dev_change_name(struct net_device *, const char *); 3698 int dev_set_alias(struct net_device *, const char *, size_t); 3699 int dev_get_alias(const struct net_device *, char *, size_t); 3700 int dev_change_net_namespace(struct net_device *, struct net *, const char *); 3701 int __dev_set_mtu(struct net_device *, int); 3702 int dev_set_mtu_ext(struct net_device *dev, int mtu, 3703 struct netlink_ext_ack *extack); 3704 int dev_set_mtu(struct net_device *, int); 3705 int dev_change_tx_queue_len(struct net_device *, unsigned long); 3706 void dev_set_group(struct net_device *, int); 3707 int dev_pre_changeaddr_notify(struct net_device *dev, const char *addr, 3708 struct netlink_ext_ack *extack); 3709 int dev_set_mac_address(struct net_device *dev, struct sockaddr *sa, 3710 struct netlink_ext_ack *extack); 3711 int dev_change_carrier(struct net_device *, bool new_carrier); 3712 int dev_get_phys_port_id(struct net_device *dev, 3713 struct netdev_phys_item_id *ppid); 3714 int dev_get_phys_port_name(struct net_device *dev, 3715 char *name, size_t len); 3716 int dev_get_port_parent_id(struct net_device *dev, 3717 struct netdev_phys_item_id *ppid, bool recurse); 3718 bool netdev_port_same_parent_id(struct net_device *a, struct net_device *b); 3719 int dev_change_proto_down(struct net_device *dev, bool proto_down); 3720 int dev_change_proto_down_generic(struct net_device *dev, bool proto_down); 3721 struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev, bool *again); 3722 struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev, 3723 struct netdev_queue *txq, int *ret); 3724 3725 typedef int (*bpf_op_t)(struct net_device *dev, struct netdev_bpf *bpf); 3726 int dev_change_xdp_fd(struct net_device *dev, struct netlink_ext_ack *extack, 3727 int fd, u32 flags); 3728 u32 __dev_xdp_query(struct net_device *dev, bpf_op_t xdp_op, 3729 enum bpf_netdev_command cmd); 3730 int xdp_umem_query(struct net_device *dev, u16 queue_id); 3731 3732 int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb); 3733 int dev_forward_skb(struct net_device *dev, struct sk_buff *skb); 3734 bool is_skb_forwardable(const struct net_device *dev, 3735 const struct sk_buff *skb); 3736 3737 static __always_inline int ____dev_forward_skb(struct net_device *dev, 3738 struct sk_buff *skb) 3739 { 3740 if (skb_orphan_frags(skb, GFP_ATOMIC) || 3741 unlikely(!is_skb_forwardable(dev, skb))) { 3742 atomic_long_inc(&dev->rx_dropped); 3743 kfree_skb(skb); 3744 return NET_RX_DROP; 3745 } 3746 3747 skb_scrub_packet(skb, true); 3748 skb->priority = 0; 3749 return 0; 3750 } 3751 3752 bool dev_nit_active(struct net_device *dev); 3753 void dev_queue_xmit_nit(struct sk_buff *skb, struct net_device *dev); 3754 3755 extern int netdev_budget; 3756 extern unsigned int netdev_budget_usecs; 3757 3758 /* Called by rtnetlink.c:rtnl_unlock() */ 3759 void netdev_run_todo(void); 3760 3761 /** 3762 * dev_put - release reference to device 3763 * @dev: network device 3764 * 3765 * Release reference to device to allow it to be freed. 3766 */ 3767 static inline void dev_put(struct net_device *dev) 3768 { 3769 this_cpu_dec(*dev->pcpu_refcnt); 3770 } 3771 3772 /** 3773 * dev_hold - get reference to device 3774 * @dev: network device 3775 * 3776 * Hold reference to device to keep it from being freed. 3777 */ 3778 static inline void dev_hold(struct net_device *dev) 3779 { 3780 this_cpu_inc(*dev->pcpu_refcnt); 3781 } 3782 3783 /* Carrier loss detection, dial on demand. The functions netif_carrier_on 3784 * and _off may be called from IRQ context, but it is caller 3785 * who is responsible for serialization of these calls. 3786 * 3787 * The name carrier is inappropriate, these functions should really be 3788 * called netif_lowerlayer_*() because they represent the state of any 3789 * kind of lower layer not just hardware media. 3790 */ 3791 3792 void linkwatch_init_dev(struct net_device *dev); 3793 void linkwatch_fire_event(struct net_device *dev); 3794 void linkwatch_forget_dev(struct net_device *dev); 3795 3796 /** 3797 * netif_carrier_ok - test if carrier present 3798 * @dev: network device 3799 * 3800 * Check if carrier is present on device 3801 */ 3802 static inline bool netif_carrier_ok(const struct net_device *dev) 3803 { 3804 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state); 3805 } 3806 3807 unsigned long dev_trans_start(struct net_device *dev); 3808 3809 void __netdev_watchdog_up(struct net_device *dev); 3810 3811 void netif_carrier_on(struct net_device *dev); 3812 3813 void netif_carrier_off(struct net_device *dev); 3814 3815 /** 3816 * netif_dormant_on - mark device as dormant. 3817 * @dev: network device 3818 * 3819 * Mark device as dormant (as per RFC2863). 3820 * 3821 * The dormant state indicates that the relevant interface is not 3822 * actually in a condition to pass packets (i.e., it is not 'up') but is 3823 * in a "pending" state, waiting for some external event. For "on- 3824 * demand" interfaces, this new state identifies the situation where the 3825 * interface is waiting for events to place it in the up state. 3826 */ 3827 static inline void netif_dormant_on(struct net_device *dev) 3828 { 3829 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state)) 3830 linkwatch_fire_event(dev); 3831 } 3832 3833 /** 3834 * netif_dormant_off - set device as not dormant. 3835 * @dev: network device 3836 * 3837 * Device is not in dormant state. 3838 */ 3839 static inline void netif_dormant_off(struct net_device *dev) 3840 { 3841 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state)) 3842 linkwatch_fire_event(dev); 3843 } 3844 3845 /** 3846 * netif_dormant - test if device is dormant 3847 * @dev: network device 3848 * 3849 * Check if device is dormant. 3850 */ 3851 static inline bool netif_dormant(const struct net_device *dev) 3852 { 3853 return test_bit(__LINK_STATE_DORMANT, &dev->state); 3854 } 3855 3856 3857 /** 3858 * netif_oper_up - test if device is operational 3859 * @dev: network device 3860 * 3861 * Check if carrier is operational 3862 */ 3863 static inline bool netif_oper_up(const struct net_device *dev) 3864 { 3865 return (dev->operstate == IF_OPER_UP || 3866 dev->operstate == IF_OPER_UNKNOWN /* backward compat */); 3867 } 3868 3869 /** 3870 * netif_device_present - is device available or removed 3871 * @dev: network device 3872 * 3873 * Check if device has not been removed from system. 3874 */ 3875 static inline bool netif_device_present(struct net_device *dev) 3876 { 3877 return test_bit(__LINK_STATE_PRESENT, &dev->state); 3878 } 3879 3880 void netif_device_detach(struct net_device *dev); 3881 3882 void netif_device_attach(struct net_device *dev); 3883 3884 /* 3885 * Network interface message level settings 3886 */ 3887 3888 enum { 3889 NETIF_MSG_DRV = 0x0001, 3890 NETIF_MSG_PROBE = 0x0002, 3891 NETIF_MSG_LINK = 0x0004, 3892 NETIF_MSG_TIMER = 0x0008, 3893 NETIF_MSG_IFDOWN = 0x0010, 3894 NETIF_MSG_IFUP = 0x0020, 3895 NETIF_MSG_RX_ERR = 0x0040, 3896 NETIF_MSG_TX_ERR = 0x0080, 3897 NETIF_MSG_TX_QUEUED = 0x0100, 3898 NETIF_MSG_INTR = 0x0200, 3899 NETIF_MSG_TX_DONE = 0x0400, 3900 NETIF_MSG_RX_STATUS = 0x0800, 3901 NETIF_MSG_PKTDATA = 0x1000, 3902 NETIF_MSG_HW = 0x2000, 3903 NETIF_MSG_WOL = 0x4000, 3904 }; 3905 3906 #define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV) 3907 #define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE) 3908 #define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK) 3909 #define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER) 3910 #define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN) 3911 #define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP) 3912 #define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR) 3913 #define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR) 3914 #define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED) 3915 #define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR) 3916 #define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE) 3917 #define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS) 3918 #define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA) 3919 #define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW) 3920 #define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL) 3921 3922 static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits) 3923 { 3924 /* use default */ 3925 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8)) 3926 return default_msg_enable_bits; 3927 if (debug_value == 0) /* no output */ 3928 return 0; 3929 /* set low N bits */ 3930 return (1U << debug_value) - 1; 3931 } 3932 3933 static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu) 3934 { 3935 spin_lock(&txq->_xmit_lock); 3936 txq->xmit_lock_owner = cpu; 3937 } 3938 3939 static inline bool __netif_tx_acquire(struct netdev_queue *txq) 3940 { 3941 __acquire(&txq->_xmit_lock); 3942 return true; 3943 } 3944 3945 static inline void __netif_tx_release(struct netdev_queue *txq) 3946 { 3947 __release(&txq->_xmit_lock); 3948 } 3949 3950 static inline void __netif_tx_lock_bh(struct netdev_queue *txq) 3951 { 3952 spin_lock_bh(&txq->_xmit_lock); 3953 txq->xmit_lock_owner = smp_processor_id(); 3954 } 3955 3956 static inline bool __netif_tx_trylock(struct netdev_queue *txq) 3957 { 3958 bool ok = spin_trylock(&txq->_xmit_lock); 3959 if (likely(ok)) 3960 txq->xmit_lock_owner = smp_processor_id(); 3961 return ok; 3962 } 3963 3964 static inline void __netif_tx_unlock(struct netdev_queue *txq) 3965 { 3966 txq->xmit_lock_owner = -1; 3967 spin_unlock(&txq->_xmit_lock); 3968 } 3969 3970 static inline void __netif_tx_unlock_bh(struct netdev_queue *txq) 3971 { 3972 txq->xmit_lock_owner = -1; 3973 spin_unlock_bh(&txq->_xmit_lock); 3974 } 3975 3976 static inline void txq_trans_update(struct netdev_queue *txq) 3977 { 3978 if (txq->xmit_lock_owner != -1) 3979 txq->trans_start = jiffies; 3980 } 3981 3982 /* legacy drivers only, netdev_start_xmit() sets txq->trans_start */ 3983 static inline void netif_trans_update(struct net_device *dev) 3984 { 3985 struct netdev_queue *txq = netdev_get_tx_queue(dev, 0); 3986 3987 if (txq->trans_start != jiffies) 3988 txq->trans_start = jiffies; 3989 } 3990 3991 /** 3992 * netif_tx_lock - grab network device transmit lock 3993 * @dev: network device 3994 * 3995 * Get network device transmit lock 3996 */ 3997 static inline void netif_tx_lock(struct net_device *dev) 3998 { 3999 unsigned int i; 4000 int cpu; 4001 4002 spin_lock(&dev->tx_global_lock); 4003 cpu = smp_processor_id(); 4004 for (i = 0; i < dev->num_tx_queues; i++) { 4005 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 4006 4007 /* We are the only thread of execution doing a 4008 * freeze, but we have to grab the _xmit_lock in 4009 * order to synchronize with threads which are in 4010 * the ->hard_start_xmit() handler and already 4011 * checked the frozen bit. 4012 */ 4013 __netif_tx_lock(txq, cpu); 4014 set_bit(__QUEUE_STATE_FROZEN, &txq->state); 4015 __netif_tx_unlock(txq); 4016 } 4017 } 4018 4019 static inline void netif_tx_lock_bh(struct net_device *dev) 4020 { 4021 local_bh_disable(); 4022 netif_tx_lock(dev); 4023 } 4024 4025 static inline void netif_tx_unlock(struct net_device *dev) 4026 { 4027 unsigned int i; 4028 4029 for (i = 0; i < dev->num_tx_queues; i++) { 4030 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 4031 4032 /* No need to grab the _xmit_lock here. If the 4033 * queue is not stopped for another reason, we 4034 * force a schedule. 4035 */ 4036 clear_bit(__QUEUE_STATE_FROZEN, &txq->state); 4037 netif_schedule_queue(txq); 4038 } 4039 spin_unlock(&dev->tx_global_lock); 4040 } 4041 4042 static inline void netif_tx_unlock_bh(struct net_device *dev) 4043 { 4044 netif_tx_unlock(dev); 4045 local_bh_enable(); 4046 } 4047 4048 #define HARD_TX_LOCK(dev, txq, cpu) { \ 4049 if ((dev->features & NETIF_F_LLTX) == 0) { \ 4050 __netif_tx_lock(txq, cpu); \ 4051 } else { \ 4052 __netif_tx_acquire(txq); \ 4053 } \ 4054 } 4055 4056 #define HARD_TX_TRYLOCK(dev, txq) \ 4057 (((dev->features & NETIF_F_LLTX) == 0) ? \ 4058 __netif_tx_trylock(txq) : \ 4059 __netif_tx_acquire(txq)) 4060 4061 #define HARD_TX_UNLOCK(dev, txq) { \ 4062 if ((dev->features & NETIF_F_LLTX) == 0) { \ 4063 __netif_tx_unlock(txq); \ 4064 } else { \ 4065 __netif_tx_release(txq); \ 4066 } \ 4067 } 4068 4069 static inline void netif_tx_disable(struct net_device *dev) 4070 { 4071 unsigned int i; 4072 int cpu; 4073 4074 local_bh_disable(); 4075 cpu = smp_processor_id(); 4076 for (i = 0; i < dev->num_tx_queues; i++) { 4077 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 4078 4079 __netif_tx_lock(txq, cpu); 4080 netif_tx_stop_queue(txq); 4081 __netif_tx_unlock(txq); 4082 } 4083 local_bh_enable(); 4084 } 4085 4086 static inline void netif_addr_lock(struct net_device *dev) 4087 { 4088 spin_lock(&dev->addr_list_lock); 4089 } 4090 4091 static inline void netif_addr_lock_bh(struct net_device *dev) 4092 { 4093 spin_lock_bh(&dev->addr_list_lock); 4094 } 4095 4096 static inline void netif_addr_unlock(struct net_device *dev) 4097 { 4098 spin_unlock(&dev->addr_list_lock); 4099 } 4100 4101 static inline void netif_addr_unlock_bh(struct net_device *dev) 4102 { 4103 spin_unlock_bh(&dev->addr_list_lock); 4104 } 4105 4106 /* 4107 * dev_addrs walker. Should be used only for read access. Call with 4108 * rcu_read_lock held. 4109 */ 4110 #define for_each_dev_addr(dev, ha) \ 4111 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list) 4112 4113 /* These functions live elsewhere (drivers/net/net_init.c, but related) */ 4114 4115 void ether_setup(struct net_device *dev); 4116 4117 /* Support for loadable net-drivers */ 4118 struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name, 4119 unsigned char name_assign_type, 4120 void (*setup)(struct net_device *), 4121 unsigned int txqs, unsigned int rxqs); 4122 #define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \ 4123 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1) 4124 4125 #define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \ 4126 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \ 4127 count) 4128 4129 int register_netdev(struct net_device *dev); 4130 void unregister_netdev(struct net_device *dev); 4131 4132 /* General hardware address lists handling functions */ 4133 int __hw_addr_sync(struct netdev_hw_addr_list *to_list, 4134 struct netdev_hw_addr_list *from_list, int addr_len); 4135 void __hw_addr_unsync(struct netdev_hw_addr_list *to_list, 4136 struct netdev_hw_addr_list *from_list, int addr_len); 4137 int __hw_addr_sync_dev(struct netdev_hw_addr_list *list, 4138 struct net_device *dev, 4139 int (*sync)(struct net_device *, const unsigned char *), 4140 int (*unsync)(struct net_device *, 4141 const unsigned char *)); 4142 int __hw_addr_ref_sync_dev(struct netdev_hw_addr_list *list, 4143 struct net_device *dev, 4144 int (*sync)(struct net_device *, 4145 const unsigned char *, int), 4146 int (*unsync)(struct net_device *, 4147 const unsigned char *, int)); 4148 void __hw_addr_ref_unsync_dev(struct netdev_hw_addr_list *list, 4149 struct net_device *dev, 4150 int (*unsync)(struct net_device *, 4151 const unsigned char *, int)); 4152 void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list, 4153 struct net_device *dev, 4154 int (*unsync)(struct net_device *, 4155 const unsigned char *)); 4156 void __hw_addr_init(struct netdev_hw_addr_list *list); 4157 4158 /* Functions used for device addresses handling */ 4159 int dev_addr_add(struct net_device *dev, const unsigned char *addr, 4160 unsigned char addr_type); 4161 int dev_addr_del(struct net_device *dev, const unsigned char *addr, 4162 unsigned char addr_type); 4163 void dev_addr_flush(struct net_device *dev); 4164 int dev_addr_init(struct net_device *dev); 4165 4166 /* Functions used for unicast addresses handling */ 4167 int dev_uc_add(struct net_device *dev, const unsigned char *addr); 4168 int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr); 4169 int dev_uc_del(struct net_device *dev, const unsigned char *addr); 4170 int dev_uc_sync(struct net_device *to, struct net_device *from); 4171 int dev_uc_sync_multiple(struct net_device *to, struct net_device *from); 4172 void dev_uc_unsync(struct net_device *to, struct net_device *from); 4173 void dev_uc_flush(struct net_device *dev); 4174 void dev_uc_init(struct net_device *dev); 4175 4176 /** 4177 * __dev_uc_sync - Synchonize device's unicast list 4178 * @dev: device to sync 4179 * @sync: function to call if address should be added 4180 * @unsync: function to call if address should be removed 4181 * 4182 * Add newly added addresses to the interface, and release 4183 * addresses that have been deleted. 4184 */ 4185 static inline int __dev_uc_sync(struct net_device *dev, 4186 int (*sync)(struct net_device *, 4187 const unsigned char *), 4188 int (*unsync)(struct net_device *, 4189 const unsigned char *)) 4190 { 4191 return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync); 4192 } 4193 4194 /** 4195 * __dev_uc_unsync - Remove synchronized addresses from device 4196 * @dev: device to sync 4197 * @unsync: function to call if address should be removed 4198 * 4199 * Remove all addresses that were added to the device by dev_uc_sync(). 4200 */ 4201 static inline void __dev_uc_unsync(struct net_device *dev, 4202 int (*unsync)(struct net_device *, 4203 const unsigned char *)) 4204 { 4205 __hw_addr_unsync_dev(&dev->uc, dev, unsync); 4206 } 4207 4208 /* Functions used for multicast addresses handling */ 4209 int dev_mc_add(struct net_device *dev, const unsigned char *addr); 4210 int dev_mc_add_global(struct net_device *dev, const unsigned char *addr); 4211 int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr); 4212 int dev_mc_del(struct net_device *dev, const unsigned char *addr); 4213 int dev_mc_del_global(struct net_device *dev, const unsigned char *addr); 4214 int dev_mc_sync(struct net_device *to, struct net_device *from); 4215 int dev_mc_sync_multiple(struct net_device *to, struct net_device *from); 4216 void dev_mc_unsync(struct net_device *to, struct net_device *from); 4217 void dev_mc_flush(struct net_device *dev); 4218 void dev_mc_init(struct net_device *dev); 4219 4220 /** 4221 * __dev_mc_sync - Synchonize device's multicast list 4222 * @dev: device to sync 4223 * @sync: function to call if address should be added 4224 * @unsync: function to call if address should be removed 4225 * 4226 * Add newly added addresses to the interface, and release 4227 * addresses that have been deleted. 4228 */ 4229 static inline int __dev_mc_sync(struct net_device *dev, 4230 int (*sync)(struct net_device *, 4231 const unsigned char *), 4232 int (*unsync)(struct net_device *, 4233 const unsigned char *)) 4234 { 4235 return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync); 4236 } 4237 4238 /** 4239 * __dev_mc_unsync - Remove synchronized addresses from device 4240 * @dev: device to sync 4241 * @unsync: function to call if address should be removed 4242 * 4243 * Remove all addresses that were added to the device by dev_mc_sync(). 4244 */ 4245 static inline void __dev_mc_unsync(struct net_device *dev, 4246 int (*unsync)(struct net_device *, 4247 const unsigned char *)) 4248 { 4249 __hw_addr_unsync_dev(&dev->mc, dev, unsync); 4250 } 4251 4252 /* Functions used for secondary unicast and multicast support */ 4253 void dev_set_rx_mode(struct net_device *dev); 4254 void __dev_set_rx_mode(struct net_device *dev); 4255 int dev_set_promiscuity(struct net_device *dev, int inc); 4256 int dev_set_allmulti(struct net_device *dev, int inc); 4257 void netdev_state_change(struct net_device *dev); 4258 void netdev_notify_peers(struct net_device *dev); 4259 void netdev_features_change(struct net_device *dev); 4260 /* Load a device via the kmod */ 4261 void dev_load(struct net *net, const char *name); 4262 struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev, 4263 struct rtnl_link_stats64 *storage); 4264 void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64, 4265 const struct net_device_stats *netdev_stats); 4266 4267 extern int netdev_max_backlog; 4268 extern int netdev_tstamp_prequeue; 4269 extern int weight_p; 4270 extern int dev_weight_rx_bias; 4271 extern int dev_weight_tx_bias; 4272 extern int dev_rx_weight; 4273 extern int dev_tx_weight; 4274 extern int gro_normal_batch; 4275 4276 bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev); 4277 struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev, 4278 struct list_head **iter); 4279 struct net_device *netdev_all_upper_get_next_dev_rcu(struct net_device *dev, 4280 struct list_head **iter); 4281 4282 /* iterate through upper list, must be called under RCU read lock */ 4283 #define netdev_for_each_upper_dev_rcu(dev, updev, iter) \ 4284 for (iter = &(dev)->adj_list.upper, \ 4285 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \ 4286 updev; \ 4287 updev = netdev_upper_get_next_dev_rcu(dev, &(iter))) 4288 4289 int netdev_walk_all_upper_dev_rcu(struct net_device *dev, 4290 int (*fn)(struct net_device *upper_dev, 4291 void *data), 4292 void *data); 4293 4294 bool netdev_has_upper_dev_all_rcu(struct net_device *dev, 4295 struct net_device *upper_dev); 4296 4297 bool netdev_has_any_upper_dev(struct net_device *dev); 4298 4299 void *netdev_lower_get_next_private(struct net_device *dev, 4300 struct list_head **iter); 4301 void *netdev_lower_get_next_private_rcu(struct net_device *dev, 4302 struct list_head **iter); 4303 4304 #define netdev_for_each_lower_private(dev, priv, iter) \ 4305 for (iter = (dev)->adj_list.lower.next, \ 4306 priv = netdev_lower_get_next_private(dev, &(iter)); \ 4307 priv; \ 4308 priv = netdev_lower_get_next_private(dev, &(iter))) 4309 4310 #define netdev_for_each_lower_private_rcu(dev, priv, iter) \ 4311 for (iter = &(dev)->adj_list.lower, \ 4312 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \ 4313 priv; \ 4314 priv = netdev_lower_get_next_private_rcu(dev, &(iter))) 4315 4316 void *netdev_lower_get_next(struct net_device *dev, 4317 struct list_head **iter); 4318 4319 #define netdev_for_each_lower_dev(dev, ldev, iter) \ 4320 for (iter = (dev)->adj_list.lower.next, \ 4321 ldev = netdev_lower_get_next(dev, &(iter)); \ 4322 ldev; \ 4323 ldev = netdev_lower_get_next(dev, &(iter))) 4324 4325 struct net_device *netdev_all_lower_get_next(struct net_device *dev, 4326 struct list_head **iter); 4327 struct net_device *netdev_all_lower_get_next_rcu(struct net_device *dev, 4328 struct list_head **iter); 4329 4330 int netdev_walk_all_lower_dev(struct net_device *dev, 4331 int (*fn)(struct net_device *lower_dev, 4332 void *data), 4333 void *data); 4334 int netdev_walk_all_lower_dev_rcu(struct net_device *dev, 4335 int (*fn)(struct net_device *lower_dev, 4336 void *data), 4337 void *data); 4338 4339 void *netdev_adjacent_get_private(struct list_head *adj_list); 4340 void *netdev_lower_get_first_private_rcu(struct net_device *dev); 4341 struct net_device *netdev_master_upper_dev_get(struct net_device *dev); 4342 struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev); 4343 int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev, 4344 struct netlink_ext_ack *extack); 4345 int netdev_master_upper_dev_link(struct net_device *dev, 4346 struct net_device *upper_dev, 4347 void *upper_priv, void *upper_info, 4348 struct netlink_ext_ack *extack); 4349 void netdev_upper_dev_unlink(struct net_device *dev, 4350 struct net_device *upper_dev); 4351 int netdev_adjacent_change_prepare(struct net_device *old_dev, 4352 struct net_device *new_dev, 4353 struct net_device *dev, 4354 struct netlink_ext_ack *extack); 4355 void netdev_adjacent_change_commit(struct net_device *old_dev, 4356 struct net_device *new_dev, 4357 struct net_device *dev); 4358 void netdev_adjacent_change_abort(struct net_device *old_dev, 4359 struct net_device *new_dev, 4360 struct net_device *dev); 4361 void netdev_adjacent_rename_links(struct net_device *dev, char *oldname); 4362 void *netdev_lower_dev_get_private(struct net_device *dev, 4363 struct net_device *lower_dev); 4364 void netdev_lower_state_changed(struct net_device *lower_dev, 4365 void *lower_state_info); 4366 4367 /* RSS keys are 40 or 52 bytes long */ 4368 #define NETDEV_RSS_KEY_LEN 52 4369 extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN] __read_mostly; 4370 void netdev_rss_key_fill(void *buffer, size_t len); 4371 4372 int skb_checksum_help(struct sk_buff *skb); 4373 int skb_crc32c_csum_help(struct sk_buff *skb); 4374 int skb_csum_hwoffload_help(struct sk_buff *skb, 4375 const netdev_features_t features); 4376 4377 struct sk_buff *__skb_gso_segment(struct sk_buff *skb, 4378 netdev_features_t features, bool tx_path); 4379 struct sk_buff *skb_mac_gso_segment(struct sk_buff *skb, 4380 netdev_features_t features); 4381 4382 struct netdev_bonding_info { 4383 ifslave slave; 4384 ifbond master; 4385 }; 4386 4387 struct netdev_notifier_bonding_info { 4388 struct netdev_notifier_info info; /* must be first */ 4389 struct netdev_bonding_info bonding_info; 4390 }; 4391 4392 void netdev_bonding_info_change(struct net_device *dev, 4393 struct netdev_bonding_info *bonding_info); 4394 4395 static inline 4396 struct sk_buff *skb_gso_segment(struct sk_buff *skb, netdev_features_t features) 4397 { 4398 return __skb_gso_segment(skb, features, true); 4399 } 4400 __be16 skb_network_protocol(struct sk_buff *skb, int *depth); 4401 4402 static inline bool can_checksum_protocol(netdev_features_t features, 4403 __be16 protocol) 4404 { 4405 if (protocol == htons(ETH_P_FCOE)) 4406 return !!(features & NETIF_F_FCOE_CRC); 4407 4408 /* Assume this is an IP checksum (not SCTP CRC) */ 4409 4410 if (features & NETIF_F_HW_CSUM) { 4411 /* Can checksum everything */ 4412 return true; 4413 } 4414 4415 switch (protocol) { 4416 case htons(ETH_P_IP): 4417 return !!(features & NETIF_F_IP_CSUM); 4418 case htons(ETH_P_IPV6): 4419 return !!(features & NETIF_F_IPV6_CSUM); 4420 default: 4421 return false; 4422 } 4423 } 4424 4425 #ifdef CONFIG_BUG 4426 void netdev_rx_csum_fault(struct net_device *dev, struct sk_buff *skb); 4427 #else 4428 static inline void netdev_rx_csum_fault(struct net_device *dev, 4429 struct sk_buff *skb) 4430 { 4431 } 4432 #endif 4433 /* rx skb timestamps */ 4434 void net_enable_timestamp(void); 4435 void net_disable_timestamp(void); 4436 4437 #ifdef CONFIG_PROC_FS 4438 int __init dev_proc_init(void); 4439 #else 4440 #define dev_proc_init() 0 4441 #endif 4442 4443 static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops, 4444 struct sk_buff *skb, struct net_device *dev, 4445 bool more) 4446 { 4447 __this_cpu_write(softnet_data.xmit.more, more); 4448 return ops->ndo_start_xmit(skb, dev); 4449 } 4450 4451 static inline bool netdev_xmit_more(void) 4452 { 4453 return __this_cpu_read(softnet_data.xmit.more); 4454 } 4455 4456 static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev, 4457 struct netdev_queue *txq, bool more) 4458 { 4459 const struct net_device_ops *ops = dev->netdev_ops; 4460 netdev_tx_t rc; 4461 4462 rc = __netdev_start_xmit(ops, skb, dev, more); 4463 if (rc == NETDEV_TX_OK) 4464 txq_trans_update(txq); 4465 4466 return rc; 4467 } 4468 4469 int netdev_class_create_file_ns(const struct class_attribute *class_attr, 4470 const void *ns); 4471 void netdev_class_remove_file_ns(const struct class_attribute *class_attr, 4472 const void *ns); 4473 4474 static inline int netdev_class_create_file(const struct class_attribute *class_attr) 4475 { 4476 return netdev_class_create_file_ns(class_attr, NULL); 4477 } 4478 4479 static inline void netdev_class_remove_file(const struct class_attribute *class_attr) 4480 { 4481 netdev_class_remove_file_ns(class_attr, NULL); 4482 } 4483 4484 extern const struct kobj_ns_type_operations net_ns_type_operations; 4485 4486 const char *netdev_drivername(const struct net_device *dev); 4487 4488 void linkwatch_run_queue(void); 4489 4490 static inline netdev_features_t netdev_intersect_features(netdev_features_t f1, 4491 netdev_features_t f2) 4492 { 4493 if ((f1 ^ f2) & NETIF_F_HW_CSUM) { 4494 if (f1 & NETIF_F_HW_CSUM) 4495 f1 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM); 4496 else 4497 f2 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM); 4498 } 4499 4500 return f1 & f2; 4501 } 4502 4503 static inline netdev_features_t netdev_get_wanted_features( 4504 struct net_device *dev) 4505 { 4506 return (dev->features & ~dev->hw_features) | dev->wanted_features; 4507 } 4508 netdev_features_t netdev_increment_features(netdev_features_t all, 4509 netdev_features_t one, netdev_features_t mask); 4510 4511 /* Allow TSO being used on stacked device : 4512 * Performing the GSO segmentation before last device 4513 * is a performance improvement. 4514 */ 4515 static inline netdev_features_t netdev_add_tso_features(netdev_features_t features, 4516 netdev_features_t mask) 4517 { 4518 return netdev_increment_features(features, NETIF_F_ALL_TSO, mask); 4519 } 4520 4521 int __netdev_update_features(struct net_device *dev); 4522 void netdev_update_features(struct net_device *dev); 4523 void netdev_change_features(struct net_device *dev); 4524 4525 void netif_stacked_transfer_operstate(const struct net_device *rootdev, 4526 struct net_device *dev); 4527 4528 netdev_features_t passthru_features_check(struct sk_buff *skb, 4529 struct net_device *dev, 4530 netdev_features_t features); 4531 netdev_features_t netif_skb_features(struct sk_buff *skb); 4532 4533 static inline bool net_gso_ok(netdev_features_t features, int gso_type) 4534 { 4535 netdev_features_t feature = (netdev_features_t)gso_type << NETIF_F_GSO_SHIFT; 4536 4537 /* check flags correspondence */ 4538 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT)); 4539 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT)); 4540 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT)); 4541 BUILD_BUG_ON(SKB_GSO_TCP_FIXEDID != (NETIF_F_TSO_MANGLEID >> NETIF_F_GSO_SHIFT)); 4542 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT)); 4543 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT)); 4544 BUILD_BUG_ON(SKB_GSO_GRE != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT)); 4545 BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT)); 4546 BUILD_BUG_ON(SKB_GSO_IPXIP4 != (NETIF_F_GSO_IPXIP4 >> NETIF_F_GSO_SHIFT)); 4547 BUILD_BUG_ON(SKB_GSO_IPXIP6 != (NETIF_F_GSO_IPXIP6 >> NETIF_F_GSO_SHIFT)); 4548 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT)); 4549 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT)); 4550 BUILD_BUG_ON(SKB_GSO_PARTIAL != (NETIF_F_GSO_PARTIAL >> NETIF_F_GSO_SHIFT)); 4551 BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT)); 4552 BUILD_BUG_ON(SKB_GSO_SCTP != (NETIF_F_GSO_SCTP >> NETIF_F_GSO_SHIFT)); 4553 BUILD_BUG_ON(SKB_GSO_ESP != (NETIF_F_GSO_ESP >> NETIF_F_GSO_SHIFT)); 4554 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_GSO_UDP >> NETIF_F_GSO_SHIFT)); 4555 BUILD_BUG_ON(SKB_GSO_UDP_L4 != (NETIF_F_GSO_UDP_L4 >> NETIF_F_GSO_SHIFT)); 4556 4557 return (features & feature) == feature; 4558 } 4559 4560 static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features) 4561 { 4562 return net_gso_ok(features, skb_shinfo(skb)->gso_type) && 4563 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST)); 4564 } 4565 4566 static inline bool netif_needs_gso(struct sk_buff *skb, 4567 netdev_features_t features) 4568 { 4569 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) || 4570 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) && 4571 (skb->ip_summed != CHECKSUM_UNNECESSARY))); 4572 } 4573 4574 static inline void netif_set_gso_max_size(struct net_device *dev, 4575 unsigned int size) 4576 { 4577 dev->gso_max_size = size; 4578 } 4579 4580 static inline void skb_gso_error_unwind(struct sk_buff *skb, __be16 protocol, 4581 int pulled_hlen, u16 mac_offset, 4582 int mac_len) 4583 { 4584 skb->protocol = protocol; 4585 skb->encapsulation = 1; 4586 skb_push(skb, pulled_hlen); 4587 skb_reset_transport_header(skb); 4588 skb->mac_header = mac_offset; 4589 skb->network_header = skb->mac_header + mac_len; 4590 skb->mac_len = mac_len; 4591 } 4592 4593 static inline bool netif_is_macsec(const struct net_device *dev) 4594 { 4595 return dev->priv_flags & IFF_MACSEC; 4596 } 4597 4598 static inline bool netif_is_macvlan(const struct net_device *dev) 4599 { 4600 return dev->priv_flags & IFF_MACVLAN; 4601 } 4602 4603 static inline bool netif_is_macvlan_port(const struct net_device *dev) 4604 { 4605 return dev->priv_flags & IFF_MACVLAN_PORT; 4606 } 4607 4608 static inline bool netif_is_bond_master(const struct net_device *dev) 4609 { 4610 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING; 4611 } 4612 4613 static inline bool netif_is_bond_slave(const struct net_device *dev) 4614 { 4615 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING; 4616 } 4617 4618 static inline bool netif_supports_nofcs(struct net_device *dev) 4619 { 4620 return dev->priv_flags & IFF_SUPP_NOFCS; 4621 } 4622 4623 static inline bool netif_has_l3_rx_handler(const struct net_device *dev) 4624 { 4625 return dev->priv_flags & IFF_L3MDEV_RX_HANDLER; 4626 } 4627 4628 static inline bool netif_is_l3_master(const struct net_device *dev) 4629 { 4630 return dev->priv_flags & IFF_L3MDEV_MASTER; 4631 } 4632 4633 static inline bool netif_is_l3_slave(const struct net_device *dev) 4634 { 4635 return dev->priv_flags & IFF_L3MDEV_SLAVE; 4636 } 4637 4638 static inline bool netif_is_bridge_master(const struct net_device *dev) 4639 { 4640 return dev->priv_flags & IFF_EBRIDGE; 4641 } 4642 4643 static inline bool netif_is_bridge_port(const struct net_device *dev) 4644 { 4645 return dev->priv_flags & IFF_BRIDGE_PORT; 4646 } 4647 4648 static inline bool netif_is_ovs_master(const struct net_device *dev) 4649 { 4650 return dev->priv_flags & IFF_OPENVSWITCH; 4651 } 4652 4653 static inline bool netif_is_ovs_port(const struct net_device *dev) 4654 { 4655 return dev->priv_flags & IFF_OVS_DATAPATH; 4656 } 4657 4658 static inline bool netif_is_team_master(const struct net_device *dev) 4659 { 4660 return dev->priv_flags & IFF_TEAM; 4661 } 4662 4663 static inline bool netif_is_team_port(const struct net_device *dev) 4664 { 4665 return dev->priv_flags & IFF_TEAM_PORT; 4666 } 4667 4668 static inline bool netif_is_lag_master(const struct net_device *dev) 4669 { 4670 return netif_is_bond_master(dev) || netif_is_team_master(dev); 4671 } 4672 4673 static inline bool netif_is_lag_port(const struct net_device *dev) 4674 { 4675 return netif_is_bond_slave(dev) || netif_is_team_port(dev); 4676 } 4677 4678 static inline bool netif_is_rxfh_configured(const struct net_device *dev) 4679 { 4680 return dev->priv_flags & IFF_RXFH_CONFIGURED; 4681 } 4682 4683 static inline bool netif_is_failover(const struct net_device *dev) 4684 { 4685 return dev->priv_flags & IFF_FAILOVER; 4686 } 4687 4688 static inline bool netif_is_failover_slave(const struct net_device *dev) 4689 { 4690 return dev->priv_flags & IFF_FAILOVER_SLAVE; 4691 } 4692 4693 /* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */ 4694 static inline void netif_keep_dst(struct net_device *dev) 4695 { 4696 dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM); 4697 } 4698 4699 /* return true if dev can't cope with mtu frames that need vlan tag insertion */ 4700 static inline bool netif_reduces_vlan_mtu(struct net_device *dev) 4701 { 4702 /* TODO: reserve and use an additional IFF bit, if we get more users */ 4703 return dev->priv_flags & IFF_MACSEC; 4704 } 4705 4706 extern struct pernet_operations __net_initdata loopback_net_ops; 4707 4708 /* Logging, debugging and troubleshooting/diagnostic helpers. */ 4709 4710 /* netdev_printk helpers, similar to dev_printk */ 4711 4712 static inline const char *netdev_name(const struct net_device *dev) 4713 { 4714 if (!dev->name[0] || strchr(dev->name, '%')) 4715 return "(unnamed net_device)"; 4716 return dev->name; 4717 } 4718 4719 static inline bool netdev_unregistering(const struct net_device *dev) 4720 { 4721 return dev->reg_state == NETREG_UNREGISTERING; 4722 } 4723 4724 static inline const char *netdev_reg_state(const struct net_device *dev) 4725 { 4726 switch (dev->reg_state) { 4727 case NETREG_UNINITIALIZED: return " (uninitialized)"; 4728 case NETREG_REGISTERED: return ""; 4729 case NETREG_UNREGISTERING: return " (unregistering)"; 4730 case NETREG_UNREGISTERED: return " (unregistered)"; 4731 case NETREG_RELEASED: return " (released)"; 4732 case NETREG_DUMMY: return " (dummy)"; 4733 } 4734 4735 WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, dev->reg_state); 4736 return " (unknown)"; 4737 } 4738 4739 __printf(3, 4) __cold 4740 void netdev_printk(const char *level, const struct net_device *dev, 4741 const char *format, ...); 4742 __printf(2, 3) __cold 4743 void netdev_emerg(const struct net_device *dev, const char *format, ...); 4744 __printf(2, 3) __cold 4745 void netdev_alert(const struct net_device *dev, const char *format, ...); 4746 __printf(2, 3) __cold 4747 void netdev_crit(const struct net_device *dev, const char *format, ...); 4748 __printf(2, 3) __cold 4749 void netdev_err(const struct net_device *dev, const char *format, ...); 4750 __printf(2, 3) __cold 4751 void netdev_warn(const struct net_device *dev, const char *format, ...); 4752 __printf(2, 3) __cold 4753 void netdev_notice(const struct net_device *dev, const char *format, ...); 4754 __printf(2, 3) __cold 4755 void netdev_info(const struct net_device *dev, const char *format, ...); 4756 4757 #define netdev_level_once(level, dev, fmt, ...) \ 4758 do { \ 4759 static bool __print_once __read_mostly; \ 4760 \ 4761 if (!__print_once) { \ 4762 __print_once = true; \ 4763 netdev_printk(level, dev, fmt, ##__VA_ARGS__); \ 4764 } \ 4765 } while (0) 4766 4767 #define netdev_emerg_once(dev, fmt, ...) \ 4768 netdev_level_once(KERN_EMERG, dev, fmt, ##__VA_ARGS__) 4769 #define netdev_alert_once(dev, fmt, ...) \ 4770 netdev_level_once(KERN_ALERT, dev, fmt, ##__VA_ARGS__) 4771 #define netdev_crit_once(dev, fmt, ...) \ 4772 netdev_level_once(KERN_CRIT, dev, fmt, ##__VA_ARGS__) 4773 #define netdev_err_once(dev, fmt, ...) \ 4774 netdev_level_once(KERN_ERR, dev, fmt, ##__VA_ARGS__) 4775 #define netdev_warn_once(dev, fmt, ...) \ 4776 netdev_level_once(KERN_WARNING, dev, fmt, ##__VA_ARGS__) 4777 #define netdev_notice_once(dev, fmt, ...) \ 4778 netdev_level_once(KERN_NOTICE, dev, fmt, ##__VA_ARGS__) 4779 #define netdev_info_once(dev, fmt, ...) \ 4780 netdev_level_once(KERN_INFO, dev, fmt, ##__VA_ARGS__) 4781 4782 #define MODULE_ALIAS_NETDEV(device) \ 4783 MODULE_ALIAS("netdev-" device) 4784 4785 #if defined(CONFIG_DYNAMIC_DEBUG) 4786 #define netdev_dbg(__dev, format, args...) \ 4787 do { \ 4788 dynamic_netdev_dbg(__dev, format, ##args); \ 4789 } while (0) 4790 #elif defined(DEBUG) 4791 #define netdev_dbg(__dev, format, args...) \ 4792 netdev_printk(KERN_DEBUG, __dev, format, ##args) 4793 #else 4794 #define netdev_dbg(__dev, format, args...) \ 4795 ({ \ 4796 if (0) \ 4797 netdev_printk(KERN_DEBUG, __dev, format, ##args); \ 4798 }) 4799 #endif 4800 4801 #if defined(VERBOSE_DEBUG) 4802 #define netdev_vdbg netdev_dbg 4803 #else 4804 4805 #define netdev_vdbg(dev, format, args...) \ 4806 ({ \ 4807 if (0) \ 4808 netdev_printk(KERN_DEBUG, dev, format, ##args); \ 4809 0; \ 4810 }) 4811 #endif 4812 4813 /* 4814 * netdev_WARN() acts like dev_printk(), but with the key difference 4815 * of using a WARN/WARN_ON to get the message out, including the 4816 * file/line information and a backtrace. 4817 */ 4818 #define netdev_WARN(dev, format, args...) \ 4819 WARN(1, "netdevice: %s%s: " format, netdev_name(dev), \ 4820 netdev_reg_state(dev), ##args) 4821 4822 #define netdev_WARN_ONCE(dev, format, args...) \ 4823 WARN_ONCE(1, "netdevice: %s%s: " format, netdev_name(dev), \ 4824 netdev_reg_state(dev), ##args) 4825 4826 /* netif printk helpers, similar to netdev_printk */ 4827 4828 #define netif_printk(priv, type, level, dev, fmt, args...) \ 4829 do { \ 4830 if (netif_msg_##type(priv)) \ 4831 netdev_printk(level, (dev), fmt, ##args); \ 4832 } while (0) 4833 4834 #define netif_level(level, priv, type, dev, fmt, args...) \ 4835 do { \ 4836 if (netif_msg_##type(priv)) \ 4837 netdev_##level(dev, fmt, ##args); \ 4838 } while (0) 4839 4840 #define netif_emerg(priv, type, dev, fmt, args...) \ 4841 netif_level(emerg, priv, type, dev, fmt, ##args) 4842 #define netif_alert(priv, type, dev, fmt, args...) \ 4843 netif_level(alert, priv, type, dev, fmt, ##args) 4844 #define netif_crit(priv, type, dev, fmt, args...) \ 4845 netif_level(crit, priv, type, dev, fmt, ##args) 4846 #define netif_err(priv, type, dev, fmt, args...) \ 4847 netif_level(err, priv, type, dev, fmt, ##args) 4848 #define netif_warn(priv, type, dev, fmt, args...) \ 4849 netif_level(warn, priv, type, dev, fmt, ##args) 4850 #define netif_notice(priv, type, dev, fmt, args...) \ 4851 netif_level(notice, priv, type, dev, fmt, ##args) 4852 #define netif_info(priv, type, dev, fmt, args...) \ 4853 netif_level(info, priv, type, dev, fmt, ##args) 4854 4855 #if defined(CONFIG_DYNAMIC_DEBUG) 4856 #define netif_dbg(priv, type, netdev, format, args...) \ 4857 do { \ 4858 if (netif_msg_##type(priv)) \ 4859 dynamic_netdev_dbg(netdev, format, ##args); \ 4860 } while (0) 4861 #elif defined(DEBUG) 4862 #define netif_dbg(priv, type, dev, format, args...) \ 4863 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args) 4864 #else 4865 #define netif_dbg(priv, type, dev, format, args...) \ 4866 ({ \ 4867 if (0) \ 4868 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \ 4869 0; \ 4870 }) 4871 #endif 4872 4873 /* if @cond then downgrade to debug, else print at @level */ 4874 #define netif_cond_dbg(priv, type, netdev, cond, level, fmt, args...) \ 4875 do { \ 4876 if (cond) \ 4877 netif_dbg(priv, type, netdev, fmt, ##args); \ 4878 else \ 4879 netif_ ## level(priv, type, netdev, fmt, ##args); \ 4880 } while (0) 4881 4882 #if defined(VERBOSE_DEBUG) 4883 #define netif_vdbg netif_dbg 4884 #else 4885 #define netif_vdbg(priv, type, dev, format, args...) \ 4886 ({ \ 4887 if (0) \ 4888 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \ 4889 0; \ 4890 }) 4891 #endif 4892 4893 /* 4894 * The list of packet types we will receive (as opposed to discard) 4895 * and the routines to invoke. 4896 * 4897 * Why 16. Because with 16 the only overlap we get on a hash of the 4898 * low nibble of the protocol value is RARP/SNAP/X.25. 4899 * 4900 * 0800 IP 4901 * 0001 802.3 4902 * 0002 AX.25 4903 * 0004 802.2 4904 * 8035 RARP 4905 * 0005 SNAP 4906 * 0805 X.25 4907 * 0806 ARP 4908 * 8137 IPX 4909 * 0009 Localtalk 4910 * 86DD IPv6 4911 */ 4912 #define PTYPE_HASH_SIZE (16) 4913 #define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1) 4914 4915 extern struct net_device *blackhole_netdev; 4916 4917 #endif /* _LINUX_NETDEVICE_H */ 4918