1 /* SPDX-License-Identifier: GPL-2.0-or-later */ 2 /* 3 * INET An implementation of the TCP/IP protocol suite for the LINUX 4 * operating system. INET is implemented using the BSD Socket 5 * interface as the means of communication with the user level. 6 * 7 * Definitions for the Interfaces handler. 8 * 9 * Version: @(#)dev.h 1.0.10 08/12/93 10 * 11 * Authors: Ross Biro 12 * Fred N. van Kempen, <[email protected]> 13 * Corey Minyard <[email protected]> 14 * Donald J. Becker, <[email protected]> 15 * Alan Cox, <[email protected]> 16 * Bjorn Ekwall. <[email protected]> 17 * Pekka Riikonen <[email protected]> 18 * 19 * Moved to /usr/include/linux for NET3 20 */ 21 #ifndef _LINUX_NETDEVICE_H 22 #define _LINUX_NETDEVICE_H 23 24 #include <linux/timer.h> 25 #include <linux/bug.h> 26 #include <linux/delay.h> 27 #include <linux/atomic.h> 28 #include <linux/prefetch.h> 29 #include <asm/cache.h> 30 #include <asm/byteorder.h> 31 32 #include <linux/percpu.h> 33 #include <linux/rculist.h> 34 #include <linux/workqueue.h> 35 #include <linux/dynamic_queue_limits.h> 36 37 #include <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 TC_SETUP_QDISC_ETS, 853 }; 854 855 /* These structures hold the attributes of bpf state that are being passed 856 * to the netdevice through the bpf op. 857 */ 858 enum bpf_netdev_command { 859 /* Set or clear a bpf program used in the earliest stages of packet 860 * rx. The prog will have been loaded as BPF_PROG_TYPE_XDP. The callee 861 * is responsible for calling bpf_prog_put on any old progs that are 862 * stored. In case of error, the callee need not release the new prog 863 * reference, but on success it takes ownership and must bpf_prog_put 864 * when it is no longer used. 865 */ 866 XDP_SETUP_PROG, 867 XDP_SETUP_PROG_HW, 868 XDP_QUERY_PROG, 869 XDP_QUERY_PROG_HW, 870 /* BPF program for offload callbacks, invoked at program load time. */ 871 BPF_OFFLOAD_MAP_ALLOC, 872 BPF_OFFLOAD_MAP_FREE, 873 XDP_SETUP_XSK_UMEM, 874 }; 875 876 struct bpf_prog_offload_ops; 877 struct netlink_ext_ack; 878 struct xdp_umem; 879 880 struct netdev_bpf { 881 enum bpf_netdev_command command; 882 union { 883 /* XDP_SETUP_PROG */ 884 struct { 885 u32 flags; 886 struct bpf_prog *prog; 887 struct netlink_ext_ack *extack; 888 }; 889 /* XDP_QUERY_PROG, XDP_QUERY_PROG_HW */ 890 struct { 891 u32 prog_id; 892 /* flags with which program was installed */ 893 u32 prog_flags; 894 }; 895 /* BPF_OFFLOAD_MAP_ALLOC, BPF_OFFLOAD_MAP_FREE */ 896 struct { 897 struct bpf_offloaded_map *offmap; 898 }; 899 /* XDP_SETUP_XSK_UMEM */ 900 struct { 901 struct xdp_umem *umem; 902 u16 queue_id; 903 } xsk; 904 }; 905 }; 906 907 /* Flags for ndo_xsk_wakeup. */ 908 #define XDP_WAKEUP_RX (1 << 0) 909 #define XDP_WAKEUP_TX (1 << 1) 910 911 #ifdef CONFIG_XFRM_OFFLOAD 912 struct xfrmdev_ops { 913 int (*xdo_dev_state_add) (struct xfrm_state *x); 914 void (*xdo_dev_state_delete) (struct xfrm_state *x); 915 void (*xdo_dev_state_free) (struct xfrm_state *x); 916 bool (*xdo_dev_offload_ok) (struct sk_buff *skb, 917 struct xfrm_state *x); 918 void (*xdo_dev_state_advance_esn) (struct xfrm_state *x); 919 }; 920 #endif 921 922 struct dev_ifalias { 923 struct rcu_head rcuhead; 924 char ifalias[]; 925 }; 926 927 struct devlink; 928 struct tlsdev_ops; 929 930 struct netdev_name_node { 931 struct hlist_node hlist; 932 struct list_head list; 933 struct net_device *dev; 934 const char *name; 935 }; 936 937 int netdev_name_node_alt_create(struct net_device *dev, const char *name); 938 int netdev_name_node_alt_destroy(struct net_device *dev, const char *name); 939 940 /* 941 * This structure defines the management hooks for network devices. 942 * The following hooks can be defined; unless noted otherwise, they are 943 * optional and can be filled with a null pointer. 944 * 945 * int (*ndo_init)(struct net_device *dev); 946 * This function is called once when a network device is registered. 947 * The network device can use this for any late stage initialization 948 * or semantic validation. It can fail with an error code which will 949 * be propagated back to register_netdev. 950 * 951 * void (*ndo_uninit)(struct net_device *dev); 952 * This function is called when device is unregistered or when registration 953 * fails. It is not called if init fails. 954 * 955 * int (*ndo_open)(struct net_device *dev); 956 * This function is called when a network device transitions to the up 957 * state. 958 * 959 * int (*ndo_stop)(struct net_device *dev); 960 * This function is called when a network device transitions to the down 961 * state. 962 * 963 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb, 964 * struct net_device *dev); 965 * Called when a packet needs to be transmitted. 966 * Returns NETDEV_TX_OK. Can return NETDEV_TX_BUSY, but you should stop 967 * the queue before that can happen; it's for obsolete devices and weird 968 * corner cases, but the stack really does a non-trivial amount 969 * of useless work if you return NETDEV_TX_BUSY. 970 * Required; cannot be NULL. 971 * 972 * netdev_features_t (*ndo_features_check)(struct sk_buff *skb, 973 * struct net_device *dev 974 * netdev_features_t features); 975 * Called by core transmit path to determine if device is capable of 976 * performing offload operations on a given packet. This is to give 977 * the device an opportunity to implement any restrictions that cannot 978 * be otherwise expressed by feature flags. The check is called with 979 * the set of features that the stack has calculated and it returns 980 * those the driver believes to be appropriate. 981 * 982 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb, 983 * struct net_device *sb_dev); 984 * Called to decide which queue to use when device supports multiple 985 * transmit queues. 986 * 987 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags); 988 * This function is called to allow device receiver to make 989 * changes to configuration when multicast or promiscuous is enabled. 990 * 991 * void (*ndo_set_rx_mode)(struct net_device *dev); 992 * This function is called device changes address list filtering. 993 * If driver handles unicast address filtering, it should set 994 * IFF_UNICAST_FLT in its priv_flags. 995 * 996 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr); 997 * This function is called when the Media Access Control address 998 * needs to be changed. If this interface is not defined, the 999 * MAC address can not be changed. 1000 * 1001 * int (*ndo_validate_addr)(struct net_device *dev); 1002 * Test if Media Access Control address is valid for the device. 1003 * 1004 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd); 1005 * Called when a user requests an ioctl which can't be handled by 1006 * the generic interface code. If not defined ioctls return 1007 * not supported error code. 1008 * 1009 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map); 1010 * Used to set network devices bus interface parameters. This interface 1011 * is retained for legacy reasons; new devices should use the bus 1012 * interface (PCI) for low level management. 1013 * 1014 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu); 1015 * Called when a user wants to change the Maximum Transfer Unit 1016 * of a device. 1017 * 1018 * void (*ndo_tx_timeout)(struct net_device *dev, unsigned int txqueue); 1019 * Callback used when the transmitter has not made any progress 1020 * for dev->watchdog ticks. 1021 * 1022 * void (*ndo_get_stats64)(struct net_device *dev, 1023 * struct rtnl_link_stats64 *storage); 1024 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev); 1025 * Called when a user wants to get the network device usage 1026 * statistics. Drivers must do one of the following: 1027 * 1. Define @ndo_get_stats64 to fill in a zero-initialised 1028 * rtnl_link_stats64 structure passed by the caller. 1029 * 2. Define @ndo_get_stats to update a net_device_stats structure 1030 * (which should normally be dev->stats) and return a pointer to 1031 * it. The structure may be changed asynchronously only if each 1032 * field is written atomically. 1033 * 3. Update dev->stats asynchronously and atomically, and define 1034 * neither operation. 1035 * 1036 * bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id) 1037 * Return true if this device supports offload stats of this attr_id. 1038 * 1039 * int (*ndo_get_offload_stats)(int attr_id, const struct net_device *dev, 1040 * void *attr_data) 1041 * Get statistics for offload operations by attr_id. Write it into the 1042 * attr_data pointer. 1043 * 1044 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16 vid); 1045 * If device supports VLAN filtering this function is called when a 1046 * VLAN id is registered. 1047 * 1048 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, __be16 proto, u16 vid); 1049 * If device supports VLAN filtering this function is called when a 1050 * VLAN id is unregistered. 1051 * 1052 * void (*ndo_poll_controller)(struct net_device *dev); 1053 * 1054 * SR-IOV management functions. 1055 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac); 1056 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan, 1057 * u8 qos, __be16 proto); 1058 * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate, 1059 * int max_tx_rate); 1060 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting); 1061 * int (*ndo_set_vf_trust)(struct net_device *dev, int vf, bool setting); 1062 * int (*ndo_get_vf_config)(struct net_device *dev, 1063 * int vf, struct ifla_vf_info *ivf); 1064 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state); 1065 * int (*ndo_set_vf_port)(struct net_device *dev, int vf, 1066 * struct nlattr *port[]); 1067 * 1068 * Enable or disable the VF ability to query its RSS Redirection Table and 1069 * Hash Key. This is needed since on some devices VF share this information 1070 * with PF and querying it may introduce a theoretical security risk. 1071 * int (*ndo_set_vf_rss_query_en)(struct net_device *dev, int vf, bool setting); 1072 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb); 1073 * int (*ndo_setup_tc)(struct net_device *dev, enum tc_setup_type type, 1074 * void *type_data); 1075 * Called to setup any 'tc' scheduler, classifier or action on @dev. 1076 * This is always called from the stack with the rtnl lock held and netif 1077 * tx queues stopped. This allows the netdevice to perform queue 1078 * management safely. 1079 * 1080 * Fiber Channel over Ethernet (FCoE) offload functions. 1081 * int (*ndo_fcoe_enable)(struct net_device *dev); 1082 * Called when the FCoE protocol stack wants to start using LLD for FCoE 1083 * so the underlying device can perform whatever needed configuration or 1084 * initialization to support acceleration of FCoE traffic. 1085 * 1086 * int (*ndo_fcoe_disable)(struct net_device *dev); 1087 * Called when the FCoE protocol stack wants to stop using LLD for FCoE 1088 * so the underlying device can perform whatever needed clean-ups to 1089 * stop supporting acceleration of FCoE traffic. 1090 * 1091 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid, 1092 * struct scatterlist *sgl, unsigned int sgc); 1093 * Called when the FCoE Initiator wants to initialize an I/O that 1094 * is a possible candidate for Direct Data Placement (DDP). The LLD can 1095 * perform necessary setup and returns 1 to indicate the device is set up 1096 * successfully to perform DDP on this I/O, otherwise this returns 0. 1097 * 1098 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid); 1099 * Called when the FCoE Initiator/Target is done with the DDPed I/O as 1100 * indicated by the FC exchange id 'xid', so the underlying device can 1101 * clean up and reuse resources for later DDP requests. 1102 * 1103 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid, 1104 * struct scatterlist *sgl, unsigned int sgc); 1105 * Called when the FCoE Target wants to initialize an I/O that 1106 * is a possible candidate for Direct Data Placement (DDP). The LLD can 1107 * perform necessary setup and returns 1 to indicate the device is set up 1108 * successfully to perform DDP on this I/O, otherwise this returns 0. 1109 * 1110 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev, 1111 * struct netdev_fcoe_hbainfo *hbainfo); 1112 * Called when the FCoE Protocol stack wants information on the underlying 1113 * device. This information is utilized by the FCoE protocol stack to 1114 * register attributes with Fiber Channel management service as per the 1115 * FC-GS Fabric Device Management Information(FDMI) specification. 1116 * 1117 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type); 1118 * Called when the underlying device wants to override default World Wide 1119 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own 1120 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE 1121 * protocol stack to use. 1122 * 1123 * RFS acceleration. 1124 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb, 1125 * u16 rxq_index, u32 flow_id); 1126 * Set hardware filter for RFS. rxq_index is the target queue index; 1127 * flow_id is a flow ID to be passed to rps_may_expire_flow() later. 1128 * Return the filter ID on success, or a negative error code. 1129 * 1130 * Slave management functions (for bridge, bonding, etc). 1131 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev); 1132 * Called to make another netdev an underling. 1133 * 1134 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev); 1135 * Called to release previously enslaved netdev. 1136 * 1137 * Feature/offload setting functions. 1138 * netdev_features_t (*ndo_fix_features)(struct net_device *dev, 1139 * netdev_features_t features); 1140 * Adjusts the requested feature flags according to device-specific 1141 * constraints, and returns the resulting flags. Must not modify 1142 * the device state. 1143 * 1144 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features); 1145 * Called to update device configuration to new features. Passed 1146 * feature set might be less than what was returned by ndo_fix_features()). 1147 * Must return >0 or -errno if it changed dev->features itself. 1148 * 1149 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[], 1150 * struct net_device *dev, 1151 * const unsigned char *addr, u16 vid, u16 flags, 1152 * struct netlink_ext_ack *extack); 1153 * Adds an FDB entry to dev for addr. 1154 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[], 1155 * struct net_device *dev, 1156 * const unsigned char *addr, u16 vid) 1157 * Deletes the FDB entry from dev coresponding to addr. 1158 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb, 1159 * struct net_device *dev, struct net_device *filter_dev, 1160 * int *idx) 1161 * Used to add FDB entries to dump requests. Implementers should add 1162 * entries to skb and update idx with the number of entries. 1163 * 1164 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh, 1165 * u16 flags, struct netlink_ext_ack *extack) 1166 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq, 1167 * struct net_device *dev, u32 filter_mask, 1168 * int nlflags) 1169 * int (*ndo_bridge_dellink)(struct net_device *dev, struct nlmsghdr *nlh, 1170 * u16 flags); 1171 * 1172 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier); 1173 * Called to change device carrier. Soft-devices (like dummy, team, etc) 1174 * which do not represent real hardware may define this to allow their 1175 * userspace components to manage their virtual carrier state. Devices 1176 * that determine carrier state from physical hardware properties (eg 1177 * network cables) or protocol-dependent mechanisms (eg 1178 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function. 1179 * 1180 * int (*ndo_get_phys_port_id)(struct net_device *dev, 1181 * struct netdev_phys_item_id *ppid); 1182 * Called to get ID of physical port of this device. If driver does 1183 * not implement this, it is assumed that the hw is not able to have 1184 * multiple net devices on single physical port. 1185 * 1186 * int (*ndo_get_port_parent_id)(struct net_device *dev, 1187 * struct netdev_phys_item_id *ppid) 1188 * Called to get the parent ID of the physical port of this device. 1189 * 1190 * void (*ndo_udp_tunnel_add)(struct net_device *dev, 1191 * struct udp_tunnel_info *ti); 1192 * Called by UDP tunnel to notify a driver about the UDP port and socket 1193 * address family that a UDP tunnel is listnening to. It is called only 1194 * when a new port starts listening. The operation is protected by the 1195 * RTNL. 1196 * 1197 * void (*ndo_udp_tunnel_del)(struct net_device *dev, 1198 * struct udp_tunnel_info *ti); 1199 * Called by UDP tunnel to notify the driver about a UDP port and socket 1200 * address family that the UDP tunnel is not listening to anymore. The 1201 * operation is protected by the RTNL. 1202 * 1203 * void* (*ndo_dfwd_add_station)(struct net_device *pdev, 1204 * struct net_device *dev) 1205 * Called by upper layer devices to accelerate switching or other 1206 * station functionality into hardware. 'pdev is the lowerdev 1207 * to use for the offload and 'dev' is the net device that will 1208 * back the offload. Returns a pointer to the private structure 1209 * the upper layer will maintain. 1210 * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv) 1211 * Called by upper layer device to delete the station created 1212 * by 'ndo_dfwd_add_station'. 'pdev' is the net device backing 1213 * the station and priv is the structure returned by the add 1214 * operation. 1215 * int (*ndo_set_tx_maxrate)(struct net_device *dev, 1216 * int queue_index, u32 maxrate); 1217 * Called when a user wants to set a max-rate limitation of specific 1218 * TX queue. 1219 * int (*ndo_get_iflink)(const struct net_device *dev); 1220 * Called to get the iflink value of this device. 1221 * void (*ndo_change_proto_down)(struct net_device *dev, 1222 * bool proto_down); 1223 * This function is used to pass protocol port error state information 1224 * to the switch driver. The switch driver can react to the proto_down 1225 * by doing a phys down on the associated switch port. 1226 * int (*ndo_fill_metadata_dst)(struct net_device *dev, struct sk_buff *skb); 1227 * This function is used to get egress tunnel information for given skb. 1228 * This is useful for retrieving outer tunnel header parameters while 1229 * sampling packet. 1230 * void (*ndo_set_rx_headroom)(struct net_device *dev, int needed_headroom); 1231 * This function is used to specify the headroom that the skb must 1232 * consider when allocation skb during packet reception. Setting 1233 * appropriate rx headroom value allows avoiding skb head copy on 1234 * forward. Setting a negative value resets the rx headroom to the 1235 * default value. 1236 * int (*ndo_bpf)(struct net_device *dev, struct netdev_bpf *bpf); 1237 * This function is used to set or query state related to XDP on the 1238 * netdevice and manage BPF offload. See definition of 1239 * enum bpf_netdev_command for details. 1240 * int (*ndo_xdp_xmit)(struct net_device *dev, int n, struct xdp_frame **xdp, 1241 * u32 flags); 1242 * This function is used to submit @n XDP packets for transmit on a 1243 * netdevice. Returns number of frames successfully transmitted, frames 1244 * that got dropped are freed/returned via xdp_return_frame(). 1245 * Returns negative number, means general error invoking ndo, meaning 1246 * no frames were xmit'ed and core-caller will free all frames. 1247 * int (*ndo_xsk_wakeup)(struct net_device *dev, u32 queue_id, u32 flags); 1248 * This function is used to wake up the softirq, ksoftirqd or kthread 1249 * responsible for sending and/or receiving packets on a specific 1250 * queue id bound to an AF_XDP socket. The flags field specifies if 1251 * only RX, only Tx, or both should be woken up using the flags 1252 * XDP_WAKEUP_RX and XDP_WAKEUP_TX. 1253 * struct devlink_port *(*ndo_get_devlink_port)(struct net_device *dev); 1254 * Get devlink port instance associated with a given netdev. 1255 * Called with a reference on the netdevice and devlink locks only, 1256 * rtnl_lock is not held. 1257 */ 1258 struct net_device_ops { 1259 int (*ndo_init)(struct net_device *dev); 1260 void (*ndo_uninit)(struct net_device *dev); 1261 int (*ndo_open)(struct net_device *dev); 1262 int (*ndo_stop)(struct net_device *dev); 1263 netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb, 1264 struct net_device *dev); 1265 netdev_features_t (*ndo_features_check)(struct sk_buff *skb, 1266 struct net_device *dev, 1267 netdev_features_t features); 1268 u16 (*ndo_select_queue)(struct net_device *dev, 1269 struct sk_buff *skb, 1270 struct net_device *sb_dev); 1271 void (*ndo_change_rx_flags)(struct net_device *dev, 1272 int flags); 1273 void (*ndo_set_rx_mode)(struct net_device *dev); 1274 int (*ndo_set_mac_address)(struct net_device *dev, 1275 void *addr); 1276 int (*ndo_validate_addr)(struct net_device *dev); 1277 int (*ndo_do_ioctl)(struct net_device *dev, 1278 struct ifreq *ifr, int cmd); 1279 int (*ndo_set_config)(struct net_device *dev, 1280 struct ifmap *map); 1281 int (*ndo_change_mtu)(struct net_device *dev, 1282 int new_mtu); 1283 int (*ndo_neigh_setup)(struct net_device *dev, 1284 struct neigh_parms *); 1285 void (*ndo_tx_timeout) (struct net_device *dev, 1286 unsigned int txqueue); 1287 1288 void (*ndo_get_stats64)(struct net_device *dev, 1289 struct rtnl_link_stats64 *storage); 1290 bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id); 1291 int (*ndo_get_offload_stats)(int attr_id, 1292 const struct net_device *dev, 1293 void *attr_data); 1294 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev); 1295 1296 int (*ndo_vlan_rx_add_vid)(struct net_device *dev, 1297 __be16 proto, u16 vid); 1298 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, 1299 __be16 proto, u16 vid); 1300 #ifdef CONFIG_NET_POLL_CONTROLLER 1301 void (*ndo_poll_controller)(struct net_device *dev); 1302 int (*ndo_netpoll_setup)(struct net_device *dev, 1303 struct netpoll_info *info); 1304 void (*ndo_netpoll_cleanup)(struct net_device *dev); 1305 #endif 1306 int (*ndo_set_vf_mac)(struct net_device *dev, 1307 int queue, u8 *mac); 1308 int (*ndo_set_vf_vlan)(struct net_device *dev, 1309 int queue, u16 vlan, 1310 u8 qos, __be16 proto); 1311 int (*ndo_set_vf_rate)(struct net_device *dev, 1312 int vf, int min_tx_rate, 1313 int max_tx_rate); 1314 int (*ndo_set_vf_spoofchk)(struct net_device *dev, 1315 int vf, bool setting); 1316 int (*ndo_set_vf_trust)(struct net_device *dev, 1317 int vf, bool setting); 1318 int (*ndo_get_vf_config)(struct net_device *dev, 1319 int vf, 1320 struct ifla_vf_info *ivf); 1321 int (*ndo_set_vf_link_state)(struct net_device *dev, 1322 int vf, int link_state); 1323 int (*ndo_get_vf_stats)(struct net_device *dev, 1324 int vf, 1325 struct ifla_vf_stats 1326 *vf_stats); 1327 int (*ndo_set_vf_port)(struct net_device *dev, 1328 int vf, 1329 struct nlattr *port[]); 1330 int (*ndo_get_vf_port)(struct net_device *dev, 1331 int vf, struct sk_buff *skb); 1332 int (*ndo_get_vf_guid)(struct net_device *dev, 1333 int vf, 1334 struct ifla_vf_guid *node_guid, 1335 struct ifla_vf_guid *port_guid); 1336 int (*ndo_set_vf_guid)(struct net_device *dev, 1337 int vf, u64 guid, 1338 int guid_type); 1339 int (*ndo_set_vf_rss_query_en)( 1340 struct net_device *dev, 1341 int vf, bool setting); 1342 int (*ndo_setup_tc)(struct net_device *dev, 1343 enum tc_setup_type type, 1344 void *type_data); 1345 #if IS_ENABLED(CONFIG_FCOE) 1346 int (*ndo_fcoe_enable)(struct net_device *dev); 1347 int (*ndo_fcoe_disable)(struct net_device *dev); 1348 int (*ndo_fcoe_ddp_setup)(struct net_device *dev, 1349 u16 xid, 1350 struct scatterlist *sgl, 1351 unsigned int sgc); 1352 int (*ndo_fcoe_ddp_done)(struct net_device *dev, 1353 u16 xid); 1354 int (*ndo_fcoe_ddp_target)(struct net_device *dev, 1355 u16 xid, 1356 struct scatterlist *sgl, 1357 unsigned int sgc); 1358 int (*ndo_fcoe_get_hbainfo)(struct net_device *dev, 1359 struct netdev_fcoe_hbainfo *hbainfo); 1360 #endif 1361 1362 #if IS_ENABLED(CONFIG_LIBFCOE) 1363 #define NETDEV_FCOE_WWNN 0 1364 #define NETDEV_FCOE_WWPN 1 1365 int (*ndo_fcoe_get_wwn)(struct net_device *dev, 1366 u64 *wwn, int type); 1367 #endif 1368 1369 #ifdef CONFIG_RFS_ACCEL 1370 int (*ndo_rx_flow_steer)(struct net_device *dev, 1371 const struct sk_buff *skb, 1372 u16 rxq_index, 1373 u32 flow_id); 1374 #endif 1375 int (*ndo_add_slave)(struct net_device *dev, 1376 struct net_device *slave_dev, 1377 struct netlink_ext_ack *extack); 1378 int (*ndo_del_slave)(struct net_device *dev, 1379 struct net_device *slave_dev); 1380 netdev_features_t (*ndo_fix_features)(struct net_device *dev, 1381 netdev_features_t features); 1382 int (*ndo_set_features)(struct net_device *dev, 1383 netdev_features_t features); 1384 int (*ndo_neigh_construct)(struct net_device *dev, 1385 struct neighbour *n); 1386 void (*ndo_neigh_destroy)(struct net_device *dev, 1387 struct neighbour *n); 1388 1389 int (*ndo_fdb_add)(struct ndmsg *ndm, 1390 struct nlattr *tb[], 1391 struct net_device *dev, 1392 const unsigned char *addr, 1393 u16 vid, 1394 u16 flags, 1395 struct netlink_ext_ack *extack); 1396 int (*ndo_fdb_del)(struct ndmsg *ndm, 1397 struct nlattr *tb[], 1398 struct net_device *dev, 1399 const unsigned char *addr, 1400 u16 vid); 1401 int (*ndo_fdb_dump)(struct sk_buff *skb, 1402 struct netlink_callback *cb, 1403 struct net_device *dev, 1404 struct net_device *filter_dev, 1405 int *idx); 1406 int (*ndo_fdb_get)(struct sk_buff *skb, 1407 struct nlattr *tb[], 1408 struct net_device *dev, 1409 const unsigned char *addr, 1410 u16 vid, u32 portid, u32 seq, 1411 struct netlink_ext_ack *extack); 1412 int (*ndo_bridge_setlink)(struct net_device *dev, 1413 struct nlmsghdr *nlh, 1414 u16 flags, 1415 struct netlink_ext_ack *extack); 1416 int (*ndo_bridge_getlink)(struct sk_buff *skb, 1417 u32 pid, u32 seq, 1418 struct net_device *dev, 1419 u32 filter_mask, 1420 int nlflags); 1421 int (*ndo_bridge_dellink)(struct net_device *dev, 1422 struct nlmsghdr *nlh, 1423 u16 flags); 1424 int (*ndo_change_carrier)(struct net_device *dev, 1425 bool new_carrier); 1426 int (*ndo_get_phys_port_id)(struct net_device *dev, 1427 struct netdev_phys_item_id *ppid); 1428 int (*ndo_get_port_parent_id)(struct net_device *dev, 1429 struct netdev_phys_item_id *ppid); 1430 int (*ndo_get_phys_port_name)(struct net_device *dev, 1431 char *name, size_t len); 1432 void (*ndo_udp_tunnel_add)(struct net_device *dev, 1433 struct udp_tunnel_info *ti); 1434 void (*ndo_udp_tunnel_del)(struct net_device *dev, 1435 struct udp_tunnel_info *ti); 1436 void* (*ndo_dfwd_add_station)(struct net_device *pdev, 1437 struct net_device *dev); 1438 void (*ndo_dfwd_del_station)(struct net_device *pdev, 1439 void *priv); 1440 1441 int (*ndo_set_tx_maxrate)(struct net_device *dev, 1442 int queue_index, 1443 u32 maxrate); 1444 int (*ndo_get_iflink)(const struct net_device *dev); 1445 int (*ndo_change_proto_down)(struct net_device *dev, 1446 bool proto_down); 1447 int (*ndo_fill_metadata_dst)(struct net_device *dev, 1448 struct sk_buff *skb); 1449 void (*ndo_set_rx_headroom)(struct net_device *dev, 1450 int needed_headroom); 1451 int (*ndo_bpf)(struct net_device *dev, 1452 struct netdev_bpf *bpf); 1453 int (*ndo_xdp_xmit)(struct net_device *dev, int n, 1454 struct xdp_frame **xdp, 1455 u32 flags); 1456 int (*ndo_xsk_wakeup)(struct net_device *dev, 1457 u32 queue_id, u32 flags); 1458 struct devlink_port * (*ndo_get_devlink_port)(struct net_device *dev); 1459 }; 1460 1461 /** 1462 * enum net_device_priv_flags - &struct net_device priv_flags 1463 * 1464 * These are the &struct net_device, they are only set internally 1465 * by drivers and used in the kernel. These flags are invisible to 1466 * userspace; this means that the order of these flags can change 1467 * during any kernel release. 1468 * 1469 * You should have a pretty good reason to be extending these flags. 1470 * 1471 * @IFF_802_1Q_VLAN: 802.1Q VLAN device 1472 * @IFF_EBRIDGE: Ethernet bridging device 1473 * @IFF_BONDING: bonding master or slave 1474 * @IFF_ISATAP: ISATAP interface (RFC4214) 1475 * @IFF_WAN_HDLC: WAN HDLC device 1476 * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to 1477 * release skb->dst 1478 * @IFF_DONT_BRIDGE: disallow bridging this ether dev 1479 * @IFF_DISABLE_NETPOLL: disable netpoll at run-time 1480 * @IFF_MACVLAN_PORT: device used as macvlan port 1481 * @IFF_BRIDGE_PORT: device used as bridge port 1482 * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port 1483 * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit 1484 * @IFF_UNICAST_FLT: Supports unicast filtering 1485 * @IFF_TEAM_PORT: device used as team port 1486 * @IFF_SUPP_NOFCS: device supports sending custom FCS 1487 * @IFF_LIVE_ADDR_CHANGE: device supports hardware address 1488 * change when it's running 1489 * @IFF_MACVLAN: Macvlan device 1490 * @IFF_XMIT_DST_RELEASE_PERM: IFF_XMIT_DST_RELEASE not taking into account 1491 * underlying stacked devices 1492 * @IFF_L3MDEV_MASTER: device is an L3 master device 1493 * @IFF_NO_QUEUE: device can run without qdisc attached 1494 * @IFF_OPENVSWITCH: device is a Open vSwitch master 1495 * @IFF_L3MDEV_SLAVE: device is enslaved to an L3 master device 1496 * @IFF_TEAM: device is a team device 1497 * @IFF_RXFH_CONFIGURED: device has had Rx Flow indirection table configured 1498 * @IFF_PHONY_HEADROOM: the headroom value is controlled by an external 1499 * entity (i.e. the master device for bridged veth) 1500 * @IFF_MACSEC: device is a MACsec device 1501 * @IFF_NO_RX_HANDLER: device doesn't support the rx_handler hook 1502 * @IFF_FAILOVER: device is a failover master device 1503 * @IFF_FAILOVER_SLAVE: device is lower dev of a failover master device 1504 * @IFF_L3MDEV_RX_HANDLER: only invoke the rx handler of L3 master device 1505 * @IFF_LIVE_RENAME_OK: rename is allowed while device is up and running 1506 */ 1507 enum netdev_priv_flags { 1508 IFF_802_1Q_VLAN = 1<<0, 1509 IFF_EBRIDGE = 1<<1, 1510 IFF_BONDING = 1<<2, 1511 IFF_ISATAP = 1<<3, 1512 IFF_WAN_HDLC = 1<<4, 1513 IFF_XMIT_DST_RELEASE = 1<<5, 1514 IFF_DONT_BRIDGE = 1<<6, 1515 IFF_DISABLE_NETPOLL = 1<<7, 1516 IFF_MACVLAN_PORT = 1<<8, 1517 IFF_BRIDGE_PORT = 1<<9, 1518 IFF_OVS_DATAPATH = 1<<10, 1519 IFF_TX_SKB_SHARING = 1<<11, 1520 IFF_UNICAST_FLT = 1<<12, 1521 IFF_TEAM_PORT = 1<<13, 1522 IFF_SUPP_NOFCS = 1<<14, 1523 IFF_LIVE_ADDR_CHANGE = 1<<15, 1524 IFF_MACVLAN = 1<<16, 1525 IFF_XMIT_DST_RELEASE_PERM = 1<<17, 1526 IFF_L3MDEV_MASTER = 1<<18, 1527 IFF_NO_QUEUE = 1<<19, 1528 IFF_OPENVSWITCH = 1<<20, 1529 IFF_L3MDEV_SLAVE = 1<<21, 1530 IFF_TEAM = 1<<22, 1531 IFF_RXFH_CONFIGURED = 1<<23, 1532 IFF_PHONY_HEADROOM = 1<<24, 1533 IFF_MACSEC = 1<<25, 1534 IFF_NO_RX_HANDLER = 1<<26, 1535 IFF_FAILOVER = 1<<27, 1536 IFF_FAILOVER_SLAVE = 1<<28, 1537 IFF_L3MDEV_RX_HANDLER = 1<<29, 1538 IFF_LIVE_RENAME_OK = 1<<30, 1539 }; 1540 1541 #define IFF_802_1Q_VLAN IFF_802_1Q_VLAN 1542 #define IFF_EBRIDGE IFF_EBRIDGE 1543 #define IFF_BONDING IFF_BONDING 1544 #define IFF_ISATAP IFF_ISATAP 1545 #define IFF_WAN_HDLC IFF_WAN_HDLC 1546 #define IFF_XMIT_DST_RELEASE IFF_XMIT_DST_RELEASE 1547 #define IFF_DONT_BRIDGE IFF_DONT_BRIDGE 1548 #define IFF_DISABLE_NETPOLL IFF_DISABLE_NETPOLL 1549 #define IFF_MACVLAN_PORT IFF_MACVLAN_PORT 1550 #define IFF_BRIDGE_PORT IFF_BRIDGE_PORT 1551 #define IFF_OVS_DATAPATH IFF_OVS_DATAPATH 1552 #define IFF_TX_SKB_SHARING IFF_TX_SKB_SHARING 1553 #define IFF_UNICAST_FLT IFF_UNICAST_FLT 1554 #define IFF_TEAM_PORT IFF_TEAM_PORT 1555 #define IFF_SUPP_NOFCS IFF_SUPP_NOFCS 1556 #define IFF_LIVE_ADDR_CHANGE IFF_LIVE_ADDR_CHANGE 1557 #define IFF_MACVLAN IFF_MACVLAN 1558 #define IFF_XMIT_DST_RELEASE_PERM IFF_XMIT_DST_RELEASE_PERM 1559 #define IFF_L3MDEV_MASTER IFF_L3MDEV_MASTER 1560 #define IFF_NO_QUEUE IFF_NO_QUEUE 1561 #define IFF_OPENVSWITCH IFF_OPENVSWITCH 1562 #define IFF_L3MDEV_SLAVE IFF_L3MDEV_SLAVE 1563 #define IFF_TEAM IFF_TEAM 1564 #define IFF_RXFH_CONFIGURED IFF_RXFH_CONFIGURED 1565 #define IFF_MACSEC IFF_MACSEC 1566 #define IFF_NO_RX_HANDLER IFF_NO_RX_HANDLER 1567 #define IFF_FAILOVER IFF_FAILOVER 1568 #define IFF_FAILOVER_SLAVE IFF_FAILOVER_SLAVE 1569 #define IFF_L3MDEV_RX_HANDLER IFF_L3MDEV_RX_HANDLER 1570 #define IFF_LIVE_RENAME_OK IFF_LIVE_RENAME_OK 1571 1572 /** 1573 * struct net_device - The DEVICE structure. 1574 * 1575 * Actually, this whole structure is a big mistake. It mixes I/O 1576 * data with strictly "high-level" data, and it has to know about 1577 * almost every data structure used in the INET module. 1578 * 1579 * @name: This is the first field of the "visible" part of this structure 1580 * (i.e. as seen by users in the "Space.c" file). It is the name 1581 * of the interface. 1582 * 1583 * @name_node: Name hashlist node 1584 * @ifalias: SNMP alias 1585 * @mem_end: Shared memory end 1586 * @mem_start: Shared memory start 1587 * @base_addr: Device I/O address 1588 * @irq: Device IRQ number 1589 * 1590 * @state: Generic network queuing layer state, see netdev_state_t 1591 * @dev_list: The global list of network devices 1592 * @napi_list: List entry used for polling NAPI devices 1593 * @unreg_list: List entry when we are unregistering the 1594 * device; see the function unregister_netdev 1595 * @close_list: List entry used when we are closing the device 1596 * @ptype_all: Device-specific packet handlers for all protocols 1597 * @ptype_specific: Device-specific, protocol-specific packet handlers 1598 * 1599 * @adj_list: Directly linked devices, like slaves for bonding 1600 * @features: Currently active device features 1601 * @hw_features: User-changeable features 1602 * 1603 * @wanted_features: User-requested features 1604 * @vlan_features: Mask of features inheritable by VLAN devices 1605 * 1606 * @hw_enc_features: Mask of features inherited by encapsulating devices 1607 * This field indicates what encapsulation 1608 * offloads the hardware is capable of doing, 1609 * and drivers will need to set them appropriately. 1610 * 1611 * @mpls_features: Mask of features inheritable by MPLS 1612 * 1613 * @ifindex: interface index 1614 * @group: The group the device belongs to 1615 * 1616 * @stats: Statistics struct, which was left as a legacy, use 1617 * rtnl_link_stats64 instead 1618 * 1619 * @rx_dropped: Dropped packets by core network, 1620 * do not use this in drivers 1621 * @tx_dropped: Dropped packets by core network, 1622 * do not use this in drivers 1623 * @rx_nohandler: nohandler dropped packets by core network on 1624 * inactive devices, do not use this in drivers 1625 * @carrier_up_count: Number of times the carrier has been up 1626 * @carrier_down_count: Number of times the carrier has been down 1627 * 1628 * @wireless_handlers: List of functions to handle Wireless Extensions, 1629 * instead of ioctl, 1630 * see <net/iw_handler.h> for details. 1631 * @wireless_data: Instance data managed by the core of wireless extensions 1632 * 1633 * @netdev_ops: Includes several pointers to callbacks, 1634 * if one wants to override the ndo_*() functions 1635 * @ethtool_ops: Management operations 1636 * @ndisc_ops: Includes callbacks for different IPv6 neighbour 1637 * discovery handling. Necessary for e.g. 6LoWPAN. 1638 * @header_ops: Includes callbacks for creating,parsing,caching,etc 1639 * of Layer 2 headers. 1640 * 1641 * @flags: Interface flags (a la BSD) 1642 * @priv_flags: Like 'flags' but invisible to userspace, 1643 * see if.h for the definitions 1644 * @gflags: Global flags ( kept as legacy ) 1645 * @padded: How much padding added by alloc_netdev() 1646 * @operstate: RFC2863 operstate 1647 * @link_mode: Mapping policy to operstate 1648 * @if_port: Selectable AUI, TP, ... 1649 * @dma: DMA channel 1650 * @mtu: Interface MTU value 1651 * @min_mtu: Interface Minimum MTU value 1652 * @max_mtu: Interface Maximum MTU value 1653 * @type: Interface hardware type 1654 * @hard_header_len: Maximum hardware header length. 1655 * @min_header_len: Minimum hardware header length 1656 * 1657 * @needed_headroom: Extra headroom the hardware may need, but not in all 1658 * cases can this be guaranteed 1659 * @needed_tailroom: Extra tailroom the hardware may need, but not in all 1660 * cases can this be guaranteed. Some cases also use 1661 * LL_MAX_HEADER instead to allocate the skb 1662 * 1663 * interface address info: 1664 * 1665 * @perm_addr: Permanent hw address 1666 * @addr_assign_type: Hw address assignment type 1667 * @addr_len: Hardware address length 1668 * @upper_level: Maximum depth level of upper devices. 1669 * @lower_level: Maximum depth level of lower devices. 1670 * @neigh_priv_len: Used in neigh_alloc() 1671 * @dev_id: Used to differentiate devices that share 1672 * the same link layer address 1673 * @dev_port: Used to differentiate devices that share 1674 * the same function 1675 * @addr_list_lock: XXX: need comments on this one 1676 * @uc_promisc: Counter that indicates promiscuous mode 1677 * has been enabled due to the need to listen to 1678 * additional unicast addresses in a device that 1679 * does not implement ndo_set_rx_mode() 1680 * @uc: unicast mac addresses 1681 * @mc: multicast mac addresses 1682 * @dev_addrs: list of device hw addresses 1683 * @queues_kset: Group of all Kobjects in the Tx and RX queues 1684 * @promiscuity: Number of times the NIC is told to work in 1685 * promiscuous mode; if it becomes 0 the NIC will 1686 * exit promiscuous mode 1687 * @allmulti: Counter, enables or disables allmulticast mode 1688 * 1689 * @vlan_info: VLAN info 1690 * @dsa_ptr: dsa specific data 1691 * @tipc_ptr: TIPC specific data 1692 * @atalk_ptr: AppleTalk link 1693 * @ip_ptr: IPv4 specific data 1694 * @dn_ptr: DECnet specific data 1695 * @ip6_ptr: IPv6 specific data 1696 * @ax25_ptr: AX.25 specific data 1697 * @ieee80211_ptr: IEEE 802.11 specific data, assign before registering 1698 * 1699 * @dev_addr: Hw address (before bcast, 1700 * because most packets are unicast) 1701 * 1702 * @_rx: Array of RX queues 1703 * @num_rx_queues: Number of RX queues 1704 * allocated at register_netdev() time 1705 * @real_num_rx_queues: Number of RX queues currently active in device 1706 * 1707 * @rx_handler: handler for received packets 1708 * @rx_handler_data: XXX: need comments on this one 1709 * @miniq_ingress: ingress/clsact qdisc specific data for 1710 * ingress processing 1711 * @ingress_queue: XXX: need comments on this one 1712 * @nf_hooks_ingress: netfilter hooks executed for ingress packets 1713 * @broadcast: hw bcast address 1714 * 1715 * @rx_cpu_rmap: CPU reverse-mapping for RX completion interrupts, 1716 * indexed by RX queue number. Assigned by driver. 1717 * This must only be set if the ndo_rx_flow_steer 1718 * operation is defined 1719 * @index_hlist: Device index hash chain 1720 * 1721 * @_tx: Array of TX queues 1722 * @num_tx_queues: Number of TX queues allocated at alloc_netdev_mq() time 1723 * @real_num_tx_queues: Number of TX queues currently active in device 1724 * @qdisc: Root qdisc from userspace point of view 1725 * @tx_queue_len: Max frames per queue allowed 1726 * @tx_global_lock: XXX: need comments on this one 1727 * 1728 * @xps_maps: XXX: need comments on this one 1729 * @miniq_egress: clsact qdisc specific data for 1730 * egress processing 1731 * @watchdog_timeo: Represents the timeout that is used by 1732 * the watchdog (see dev_watchdog()) 1733 * @watchdog_timer: List of timers 1734 * 1735 * @pcpu_refcnt: Number of references to this device 1736 * @todo_list: Delayed register/unregister 1737 * @link_watch_list: XXX: need comments on this one 1738 * 1739 * @reg_state: Register/unregister state machine 1740 * @dismantle: Device is going to be freed 1741 * @rtnl_link_state: This enum represents the phases of creating 1742 * a new link 1743 * 1744 * @needs_free_netdev: Should unregister perform free_netdev? 1745 * @priv_destructor: Called from unregister 1746 * @npinfo: XXX: need comments on this one 1747 * @nd_net: Network namespace this network device is inside 1748 * 1749 * @ml_priv: Mid-layer private 1750 * @lstats: Loopback statistics 1751 * @tstats: Tunnel statistics 1752 * @dstats: Dummy statistics 1753 * @vstats: Virtual ethernet statistics 1754 * 1755 * @garp_port: GARP 1756 * @mrp_port: MRP 1757 * 1758 * @dev: Class/net/name entry 1759 * @sysfs_groups: Space for optional device, statistics and wireless 1760 * sysfs groups 1761 * 1762 * @sysfs_rx_queue_group: Space for optional per-rx queue attributes 1763 * @rtnl_link_ops: Rtnl_link_ops 1764 * 1765 * @gso_max_size: Maximum size of generic segmentation offload 1766 * @gso_max_segs: Maximum number of segments that can be passed to the 1767 * NIC for GSO 1768 * 1769 * @dcbnl_ops: Data Center Bridging netlink ops 1770 * @num_tc: Number of traffic classes in the net device 1771 * @tc_to_txq: XXX: need comments on this one 1772 * @prio_tc_map: XXX: need comments on this one 1773 * 1774 * @fcoe_ddp_xid: Max exchange id for FCoE LRO by ddp 1775 * 1776 * @priomap: XXX: need comments on this one 1777 * @phydev: Physical device may attach itself 1778 * for hardware timestamping 1779 * @sfp_bus: attached &struct sfp_bus structure. 1780 * @qdisc_tx_busylock_key: lockdep class annotating Qdisc->busylock 1781 * spinlock 1782 * @qdisc_running_key: lockdep class annotating Qdisc->running seqcount 1783 * @qdisc_xmit_lock_key: lockdep class annotating 1784 * netdev_queue->_xmit_lock spinlock 1785 * @addr_list_lock_key: lockdep class annotating 1786 * net_device->addr_list_lock spinlock 1787 * 1788 * @proto_down: protocol port state information can be sent to the 1789 * switch driver and used to set the phys state of the 1790 * switch port. 1791 * 1792 * @wol_enabled: Wake-on-LAN is enabled 1793 * 1794 * FIXME: cleanup struct net_device such that network protocol info 1795 * moves out. 1796 */ 1797 1798 struct net_device { 1799 char name[IFNAMSIZ]; 1800 struct netdev_name_node *name_node; 1801 struct dev_ifalias __rcu *ifalias; 1802 /* 1803 * I/O specific fields 1804 * FIXME: Merge these and struct ifmap into one 1805 */ 1806 unsigned long mem_end; 1807 unsigned long mem_start; 1808 unsigned long base_addr; 1809 int irq; 1810 1811 /* 1812 * Some hardware also needs these fields (state,dev_list, 1813 * napi_list,unreg_list,close_list) but they are not 1814 * part of the usual set specified in Space.c. 1815 */ 1816 1817 unsigned long state; 1818 1819 struct list_head dev_list; 1820 struct list_head napi_list; 1821 struct list_head unreg_list; 1822 struct list_head close_list; 1823 struct list_head ptype_all; 1824 struct list_head ptype_specific; 1825 1826 struct { 1827 struct list_head upper; 1828 struct list_head lower; 1829 } adj_list; 1830 1831 netdev_features_t features; 1832 netdev_features_t hw_features; 1833 netdev_features_t wanted_features; 1834 netdev_features_t vlan_features; 1835 netdev_features_t hw_enc_features; 1836 netdev_features_t mpls_features; 1837 netdev_features_t gso_partial_features; 1838 1839 int ifindex; 1840 int group; 1841 1842 struct net_device_stats stats; 1843 1844 atomic_long_t rx_dropped; 1845 atomic_long_t tx_dropped; 1846 atomic_long_t rx_nohandler; 1847 1848 /* Stats to monitor link on/off, flapping */ 1849 atomic_t carrier_up_count; 1850 atomic_t carrier_down_count; 1851 1852 #ifdef CONFIG_WIRELESS_EXT 1853 const struct iw_handler_def *wireless_handlers; 1854 struct iw_public_data *wireless_data; 1855 #endif 1856 const struct net_device_ops *netdev_ops; 1857 const struct ethtool_ops *ethtool_ops; 1858 #ifdef CONFIG_NET_L3_MASTER_DEV 1859 const struct l3mdev_ops *l3mdev_ops; 1860 #endif 1861 #if IS_ENABLED(CONFIG_IPV6) 1862 const struct ndisc_ops *ndisc_ops; 1863 #endif 1864 1865 #ifdef CONFIG_XFRM_OFFLOAD 1866 const struct xfrmdev_ops *xfrmdev_ops; 1867 #endif 1868 1869 #if IS_ENABLED(CONFIG_TLS_DEVICE) 1870 const struct tlsdev_ops *tlsdev_ops; 1871 #endif 1872 1873 const struct header_ops *header_ops; 1874 1875 unsigned int flags; 1876 unsigned int priv_flags; 1877 1878 unsigned short gflags; 1879 unsigned short padded; 1880 1881 unsigned char operstate; 1882 unsigned char link_mode; 1883 1884 unsigned char if_port; 1885 unsigned char dma; 1886 1887 /* Note : dev->mtu is often read without holding a lock. 1888 * Writers usually hold RTNL. 1889 * It is recommended to use READ_ONCE() to annotate the reads, 1890 * and to use WRITE_ONCE() to annotate the writes. 1891 */ 1892 unsigned int mtu; 1893 unsigned int min_mtu; 1894 unsigned int max_mtu; 1895 unsigned short type; 1896 unsigned short hard_header_len; 1897 unsigned char min_header_len; 1898 1899 unsigned short needed_headroom; 1900 unsigned short needed_tailroom; 1901 1902 /* Interface address info. */ 1903 unsigned char perm_addr[MAX_ADDR_LEN]; 1904 unsigned char addr_assign_type; 1905 unsigned char addr_len; 1906 unsigned char upper_level; 1907 unsigned char lower_level; 1908 unsigned short neigh_priv_len; 1909 unsigned short dev_id; 1910 unsigned short dev_port; 1911 spinlock_t addr_list_lock; 1912 unsigned char name_assign_type; 1913 bool uc_promisc; 1914 struct netdev_hw_addr_list uc; 1915 struct netdev_hw_addr_list mc; 1916 struct netdev_hw_addr_list dev_addrs; 1917 1918 #ifdef CONFIG_SYSFS 1919 struct kset *queues_kset; 1920 #endif 1921 unsigned int promiscuity; 1922 unsigned int allmulti; 1923 1924 1925 /* Protocol-specific pointers */ 1926 1927 #if IS_ENABLED(CONFIG_VLAN_8021Q) 1928 struct vlan_info __rcu *vlan_info; 1929 #endif 1930 #if IS_ENABLED(CONFIG_NET_DSA) 1931 struct dsa_port *dsa_ptr; 1932 #endif 1933 #if IS_ENABLED(CONFIG_TIPC) 1934 struct tipc_bearer __rcu *tipc_ptr; 1935 #endif 1936 #if IS_ENABLED(CONFIG_IRDA) || IS_ENABLED(CONFIG_ATALK) 1937 void *atalk_ptr; 1938 #endif 1939 struct in_device __rcu *ip_ptr; 1940 #if IS_ENABLED(CONFIG_DECNET) 1941 struct dn_dev __rcu *dn_ptr; 1942 #endif 1943 struct inet6_dev __rcu *ip6_ptr; 1944 #if IS_ENABLED(CONFIG_AX25) 1945 void *ax25_ptr; 1946 #endif 1947 struct wireless_dev *ieee80211_ptr; 1948 struct wpan_dev *ieee802154_ptr; 1949 #if IS_ENABLED(CONFIG_MPLS_ROUTING) 1950 struct mpls_dev __rcu *mpls_ptr; 1951 #endif 1952 1953 /* 1954 * Cache lines mostly used on receive path (including eth_type_trans()) 1955 */ 1956 /* Interface address info used in eth_type_trans() */ 1957 unsigned char *dev_addr; 1958 1959 struct netdev_rx_queue *_rx; 1960 unsigned int num_rx_queues; 1961 unsigned int real_num_rx_queues; 1962 1963 struct bpf_prog __rcu *xdp_prog; 1964 unsigned long gro_flush_timeout; 1965 rx_handler_func_t __rcu *rx_handler; 1966 void __rcu *rx_handler_data; 1967 1968 #ifdef CONFIG_NET_CLS_ACT 1969 struct mini_Qdisc __rcu *miniq_ingress; 1970 #endif 1971 struct netdev_queue __rcu *ingress_queue; 1972 #ifdef CONFIG_NETFILTER_INGRESS 1973 struct nf_hook_entries __rcu *nf_hooks_ingress; 1974 #endif 1975 1976 unsigned char broadcast[MAX_ADDR_LEN]; 1977 #ifdef CONFIG_RFS_ACCEL 1978 struct cpu_rmap *rx_cpu_rmap; 1979 #endif 1980 struct hlist_node index_hlist; 1981 1982 /* 1983 * Cache lines mostly used on transmit path 1984 */ 1985 struct netdev_queue *_tx ____cacheline_aligned_in_smp; 1986 unsigned int num_tx_queues; 1987 unsigned int real_num_tx_queues; 1988 struct Qdisc *qdisc; 1989 #ifdef CONFIG_NET_SCHED 1990 DECLARE_HASHTABLE (qdisc_hash, 4); 1991 #endif 1992 unsigned int tx_queue_len; 1993 spinlock_t tx_global_lock; 1994 int watchdog_timeo; 1995 1996 #ifdef CONFIG_XPS 1997 struct xps_dev_maps __rcu *xps_cpus_map; 1998 struct xps_dev_maps __rcu *xps_rxqs_map; 1999 #endif 2000 #ifdef CONFIG_NET_CLS_ACT 2001 struct mini_Qdisc __rcu *miniq_egress; 2002 #endif 2003 2004 /* These may be needed for future network-power-down code. */ 2005 struct timer_list watchdog_timer; 2006 2007 int __percpu *pcpu_refcnt; 2008 struct list_head todo_list; 2009 2010 struct list_head link_watch_list; 2011 2012 enum { NETREG_UNINITIALIZED=0, 2013 NETREG_REGISTERED, /* completed register_netdevice */ 2014 NETREG_UNREGISTERING, /* called unregister_netdevice */ 2015 NETREG_UNREGISTERED, /* completed unregister todo */ 2016 NETREG_RELEASED, /* called free_netdev */ 2017 NETREG_DUMMY, /* dummy device for NAPI poll */ 2018 } reg_state:8; 2019 2020 bool dismantle; 2021 2022 enum { 2023 RTNL_LINK_INITIALIZED, 2024 RTNL_LINK_INITIALIZING, 2025 } rtnl_link_state:16; 2026 2027 bool needs_free_netdev; 2028 void (*priv_destructor)(struct net_device *dev); 2029 2030 #ifdef CONFIG_NETPOLL 2031 struct netpoll_info __rcu *npinfo; 2032 #endif 2033 2034 possible_net_t nd_net; 2035 2036 /* mid-layer private */ 2037 union { 2038 void *ml_priv; 2039 struct pcpu_lstats __percpu *lstats; 2040 struct pcpu_sw_netstats __percpu *tstats; 2041 struct pcpu_dstats __percpu *dstats; 2042 }; 2043 2044 #if IS_ENABLED(CONFIG_GARP) 2045 struct garp_port __rcu *garp_port; 2046 #endif 2047 #if IS_ENABLED(CONFIG_MRP) 2048 struct mrp_port __rcu *mrp_port; 2049 #endif 2050 2051 struct device dev; 2052 const struct attribute_group *sysfs_groups[4]; 2053 const struct attribute_group *sysfs_rx_queue_group; 2054 2055 const struct rtnl_link_ops *rtnl_link_ops; 2056 2057 /* for setting kernel sock attribute on TCP connection setup */ 2058 #define GSO_MAX_SIZE 65536 2059 unsigned int gso_max_size; 2060 #define GSO_MAX_SEGS 65535 2061 u16 gso_max_segs; 2062 2063 #ifdef CONFIG_DCB 2064 const struct dcbnl_rtnl_ops *dcbnl_ops; 2065 #endif 2066 s16 num_tc; 2067 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE]; 2068 u8 prio_tc_map[TC_BITMASK + 1]; 2069 2070 #if IS_ENABLED(CONFIG_FCOE) 2071 unsigned int fcoe_ddp_xid; 2072 #endif 2073 #if IS_ENABLED(CONFIG_CGROUP_NET_PRIO) 2074 struct netprio_map __rcu *priomap; 2075 #endif 2076 struct phy_device *phydev; 2077 struct sfp_bus *sfp_bus; 2078 struct lock_class_key qdisc_tx_busylock_key; 2079 struct lock_class_key qdisc_running_key; 2080 struct lock_class_key qdisc_xmit_lock_key; 2081 struct lock_class_key addr_list_lock_key; 2082 bool proto_down; 2083 unsigned wol_enabled:1; 2084 }; 2085 #define to_net_dev(d) container_of(d, struct net_device, dev) 2086 2087 static inline bool netif_elide_gro(const struct net_device *dev) 2088 { 2089 if (!(dev->features & NETIF_F_GRO) || dev->xdp_prog) 2090 return true; 2091 return false; 2092 } 2093 2094 #define NETDEV_ALIGN 32 2095 2096 static inline 2097 int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio) 2098 { 2099 return dev->prio_tc_map[prio & TC_BITMASK]; 2100 } 2101 2102 static inline 2103 int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc) 2104 { 2105 if (tc >= dev->num_tc) 2106 return -EINVAL; 2107 2108 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK; 2109 return 0; 2110 } 2111 2112 int netdev_txq_to_tc(struct net_device *dev, unsigned int txq); 2113 void netdev_reset_tc(struct net_device *dev); 2114 int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset); 2115 int netdev_set_num_tc(struct net_device *dev, u8 num_tc); 2116 2117 static inline 2118 int netdev_get_num_tc(struct net_device *dev) 2119 { 2120 return dev->num_tc; 2121 } 2122 2123 void netdev_unbind_sb_channel(struct net_device *dev, 2124 struct net_device *sb_dev); 2125 int netdev_bind_sb_channel_queue(struct net_device *dev, 2126 struct net_device *sb_dev, 2127 u8 tc, u16 count, u16 offset); 2128 int netdev_set_sb_channel(struct net_device *dev, u16 channel); 2129 static inline int netdev_get_sb_channel(struct net_device *dev) 2130 { 2131 return max_t(int, -dev->num_tc, 0); 2132 } 2133 2134 static inline 2135 struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev, 2136 unsigned int index) 2137 { 2138 return &dev->_tx[index]; 2139 } 2140 2141 static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev, 2142 const struct sk_buff *skb) 2143 { 2144 return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb)); 2145 } 2146 2147 static inline void netdev_for_each_tx_queue(struct net_device *dev, 2148 void (*f)(struct net_device *, 2149 struct netdev_queue *, 2150 void *), 2151 void *arg) 2152 { 2153 unsigned int i; 2154 2155 for (i = 0; i < dev->num_tx_queues; i++) 2156 f(dev, &dev->_tx[i], arg); 2157 } 2158 2159 u16 netdev_pick_tx(struct net_device *dev, struct sk_buff *skb, 2160 struct net_device *sb_dev); 2161 struct netdev_queue *netdev_core_pick_tx(struct net_device *dev, 2162 struct sk_buff *skb, 2163 struct net_device *sb_dev); 2164 2165 /* returns the headroom that the master device needs to take in account 2166 * when forwarding to this dev 2167 */ 2168 static inline unsigned netdev_get_fwd_headroom(struct net_device *dev) 2169 { 2170 return dev->priv_flags & IFF_PHONY_HEADROOM ? 0 : dev->needed_headroom; 2171 } 2172 2173 static inline void netdev_set_rx_headroom(struct net_device *dev, int new_hr) 2174 { 2175 if (dev->netdev_ops->ndo_set_rx_headroom) 2176 dev->netdev_ops->ndo_set_rx_headroom(dev, new_hr); 2177 } 2178 2179 /* set the device rx headroom to the dev's default */ 2180 static inline void netdev_reset_rx_headroom(struct net_device *dev) 2181 { 2182 netdev_set_rx_headroom(dev, -1); 2183 } 2184 2185 /* 2186 * Net namespace inlines 2187 */ 2188 static inline 2189 struct net *dev_net(const struct net_device *dev) 2190 { 2191 return read_pnet(&dev->nd_net); 2192 } 2193 2194 static inline 2195 void dev_net_set(struct net_device *dev, struct net *net) 2196 { 2197 write_pnet(&dev->nd_net, net); 2198 } 2199 2200 /** 2201 * netdev_priv - access network device private data 2202 * @dev: network device 2203 * 2204 * Get network device private data 2205 */ 2206 static inline void *netdev_priv(const struct net_device *dev) 2207 { 2208 return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN); 2209 } 2210 2211 /* Set the sysfs physical device reference for the network logical device 2212 * if set prior to registration will cause a symlink during initialization. 2213 */ 2214 #define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev)) 2215 2216 /* Set the sysfs device type for the network logical device to allow 2217 * fine-grained identification of different network device types. For 2218 * example Ethernet, Wireless LAN, Bluetooth, WiMAX etc. 2219 */ 2220 #define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype)) 2221 2222 /* Default NAPI poll() weight 2223 * Device drivers are strongly advised to not use bigger value 2224 */ 2225 #define NAPI_POLL_WEIGHT 64 2226 2227 /** 2228 * netif_napi_add - initialize a NAPI context 2229 * @dev: network device 2230 * @napi: NAPI context 2231 * @poll: polling function 2232 * @weight: default weight 2233 * 2234 * netif_napi_add() must be used to initialize a NAPI context prior to calling 2235 * *any* of the other NAPI-related functions. 2236 */ 2237 void netif_napi_add(struct net_device *dev, struct napi_struct *napi, 2238 int (*poll)(struct napi_struct *, int), int weight); 2239 2240 /** 2241 * netif_tx_napi_add - initialize a NAPI context 2242 * @dev: network device 2243 * @napi: NAPI context 2244 * @poll: polling function 2245 * @weight: default weight 2246 * 2247 * This variant of netif_napi_add() should be used from drivers using NAPI 2248 * to exclusively poll a TX queue. 2249 * This will avoid we add it into napi_hash[], thus polluting this hash table. 2250 */ 2251 static inline void netif_tx_napi_add(struct net_device *dev, 2252 struct napi_struct *napi, 2253 int (*poll)(struct napi_struct *, int), 2254 int weight) 2255 { 2256 set_bit(NAPI_STATE_NO_BUSY_POLL, &napi->state); 2257 netif_napi_add(dev, napi, poll, weight); 2258 } 2259 2260 /** 2261 * netif_napi_del - remove a NAPI context 2262 * @napi: NAPI context 2263 * 2264 * netif_napi_del() removes a NAPI context from the network device NAPI list 2265 */ 2266 void netif_napi_del(struct napi_struct *napi); 2267 2268 struct napi_gro_cb { 2269 /* Virtual address of skb_shinfo(skb)->frags[0].page + offset. */ 2270 void *frag0; 2271 2272 /* Length of frag0. */ 2273 unsigned int frag0_len; 2274 2275 /* This indicates where we are processing relative to skb->data. */ 2276 int data_offset; 2277 2278 /* This is non-zero if the packet cannot be merged with the new skb. */ 2279 u16 flush; 2280 2281 /* Save the IP ID here and check when we get to the transport layer */ 2282 u16 flush_id; 2283 2284 /* Number of segments aggregated. */ 2285 u16 count; 2286 2287 /* Start offset for remote checksum offload */ 2288 u16 gro_remcsum_start; 2289 2290 /* jiffies when first packet was created/queued */ 2291 unsigned long age; 2292 2293 /* Used in ipv6_gro_receive() and foo-over-udp */ 2294 u16 proto; 2295 2296 /* This is non-zero if the packet may be of the same flow. */ 2297 u8 same_flow:1; 2298 2299 /* Used in tunnel GRO receive */ 2300 u8 encap_mark:1; 2301 2302 /* GRO checksum is valid */ 2303 u8 csum_valid:1; 2304 2305 /* Number of checksums via CHECKSUM_UNNECESSARY */ 2306 u8 csum_cnt:3; 2307 2308 /* Free the skb? */ 2309 u8 free:2; 2310 #define NAPI_GRO_FREE 1 2311 #define NAPI_GRO_FREE_STOLEN_HEAD 2 2312 2313 /* Used in foo-over-udp, set in udp[46]_gro_receive */ 2314 u8 is_ipv6:1; 2315 2316 /* Used in GRE, set in fou/gue_gro_receive */ 2317 u8 is_fou:1; 2318 2319 /* Used to determine if flush_id can be ignored */ 2320 u8 is_atomic:1; 2321 2322 /* Number of gro_receive callbacks this packet already went through */ 2323 u8 recursion_counter:4; 2324 2325 /* 1 bit hole */ 2326 2327 /* used to support CHECKSUM_COMPLETE for tunneling protocols */ 2328 __wsum csum; 2329 2330 /* used in skb_gro_receive() slow path */ 2331 struct sk_buff *last; 2332 }; 2333 2334 #define NAPI_GRO_CB(skb) ((struct napi_gro_cb *)(skb)->cb) 2335 2336 #define GRO_RECURSION_LIMIT 15 2337 static inline int gro_recursion_inc_test(struct sk_buff *skb) 2338 { 2339 return ++NAPI_GRO_CB(skb)->recursion_counter == GRO_RECURSION_LIMIT; 2340 } 2341 2342 typedef struct sk_buff *(*gro_receive_t)(struct list_head *, struct sk_buff *); 2343 static inline struct sk_buff *call_gro_receive(gro_receive_t cb, 2344 struct list_head *head, 2345 struct sk_buff *skb) 2346 { 2347 if (unlikely(gro_recursion_inc_test(skb))) { 2348 NAPI_GRO_CB(skb)->flush |= 1; 2349 return NULL; 2350 } 2351 2352 return cb(head, skb); 2353 } 2354 2355 typedef struct sk_buff *(*gro_receive_sk_t)(struct sock *, struct list_head *, 2356 struct sk_buff *); 2357 static inline struct sk_buff *call_gro_receive_sk(gro_receive_sk_t cb, 2358 struct sock *sk, 2359 struct list_head *head, 2360 struct sk_buff *skb) 2361 { 2362 if (unlikely(gro_recursion_inc_test(skb))) { 2363 NAPI_GRO_CB(skb)->flush |= 1; 2364 return NULL; 2365 } 2366 2367 return cb(sk, head, skb); 2368 } 2369 2370 struct packet_type { 2371 __be16 type; /* This is really htons(ether_type). */ 2372 bool ignore_outgoing; 2373 struct net_device *dev; /* NULL is wildcarded here */ 2374 int (*func) (struct sk_buff *, 2375 struct net_device *, 2376 struct packet_type *, 2377 struct net_device *); 2378 void (*list_func) (struct list_head *, 2379 struct packet_type *, 2380 struct net_device *); 2381 bool (*id_match)(struct packet_type *ptype, 2382 struct sock *sk); 2383 void *af_packet_priv; 2384 struct list_head list; 2385 }; 2386 2387 struct offload_callbacks { 2388 struct sk_buff *(*gso_segment)(struct sk_buff *skb, 2389 netdev_features_t features); 2390 struct sk_buff *(*gro_receive)(struct list_head *head, 2391 struct sk_buff *skb); 2392 int (*gro_complete)(struct sk_buff *skb, int nhoff); 2393 }; 2394 2395 struct packet_offload { 2396 __be16 type; /* This is really htons(ether_type). */ 2397 u16 priority; 2398 struct offload_callbacks callbacks; 2399 struct list_head list; 2400 }; 2401 2402 /* often modified stats are per-CPU, other are shared (netdev->stats) */ 2403 struct pcpu_sw_netstats { 2404 u64 rx_packets; 2405 u64 rx_bytes; 2406 u64 tx_packets; 2407 u64 tx_bytes; 2408 struct u64_stats_sync syncp; 2409 } __aligned(4 * sizeof(u64)); 2410 2411 struct pcpu_lstats { 2412 u64_stats_t packets; 2413 u64_stats_t bytes; 2414 struct u64_stats_sync syncp; 2415 } __aligned(2 * sizeof(u64)); 2416 2417 void dev_lstats_read(struct net_device *dev, u64 *packets, u64 *bytes); 2418 2419 static inline void dev_lstats_add(struct net_device *dev, unsigned int len) 2420 { 2421 struct pcpu_lstats *lstats = this_cpu_ptr(dev->lstats); 2422 2423 u64_stats_update_begin(&lstats->syncp); 2424 u64_stats_add(&lstats->bytes, len); 2425 u64_stats_inc(&lstats->packets); 2426 u64_stats_update_end(&lstats->syncp); 2427 } 2428 2429 #define __netdev_alloc_pcpu_stats(type, gfp) \ 2430 ({ \ 2431 typeof(type) __percpu *pcpu_stats = alloc_percpu_gfp(type, gfp);\ 2432 if (pcpu_stats) { \ 2433 int __cpu; \ 2434 for_each_possible_cpu(__cpu) { \ 2435 typeof(type) *stat; \ 2436 stat = per_cpu_ptr(pcpu_stats, __cpu); \ 2437 u64_stats_init(&stat->syncp); \ 2438 } \ 2439 } \ 2440 pcpu_stats; \ 2441 }) 2442 2443 #define netdev_alloc_pcpu_stats(type) \ 2444 __netdev_alloc_pcpu_stats(type, GFP_KERNEL) 2445 2446 enum netdev_lag_tx_type { 2447 NETDEV_LAG_TX_TYPE_UNKNOWN, 2448 NETDEV_LAG_TX_TYPE_RANDOM, 2449 NETDEV_LAG_TX_TYPE_BROADCAST, 2450 NETDEV_LAG_TX_TYPE_ROUNDROBIN, 2451 NETDEV_LAG_TX_TYPE_ACTIVEBACKUP, 2452 NETDEV_LAG_TX_TYPE_HASH, 2453 }; 2454 2455 enum netdev_lag_hash { 2456 NETDEV_LAG_HASH_NONE, 2457 NETDEV_LAG_HASH_L2, 2458 NETDEV_LAG_HASH_L34, 2459 NETDEV_LAG_HASH_L23, 2460 NETDEV_LAG_HASH_E23, 2461 NETDEV_LAG_HASH_E34, 2462 NETDEV_LAG_HASH_UNKNOWN, 2463 }; 2464 2465 struct netdev_lag_upper_info { 2466 enum netdev_lag_tx_type tx_type; 2467 enum netdev_lag_hash hash_type; 2468 }; 2469 2470 struct netdev_lag_lower_state_info { 2471 u8 link_up : 1, 2472 tx_enabled : 1; 2473 }; 2474 2475 #include <linux/notifier.h> 2476 2477 /* netdevice notifier chain. Please remember to update netdev_cmd_to_name() 2478 * and the rtnetlink notification exclusion list in rtnetlink_event() when 2479 * adding new types. 2480 */ 2481 enum netdev_cmd { 2482 NETDEV_UP = 1, /* For now you can't veto a device up/down */ 2483 NETDEV_DOWN, 2484 NETDEV_REBOOT, /* Tell a protocol stack a network interface 2485 detected a hardware crash and restarted 2486 - we can use this eg to kick tcp sessions 2487 once done */ 2488 NETDEV_CHANGE, /* Notify device state change */ 2489 NETDEV_REGISTER, 2490 NETDEV_UNREGISTER, 2491 NETDEV_CHANGEMTU, /* notify after mtu change happened */ 2492 NETDEV_CHANGEADDR, /* notify after the address change */ 2493 NETDEV_PRE_CHANGEADDR, /* notify before the address change */ 2494 NETDEV_GOING_DOWN, 2495 NETDEV_CHANGENAME, 2496 NETDEV_FEAT_CHANGE, 2497 NETDEV_BONDING_FAILOVER, 2498 NETDEV_PRE_UP, 2499 NETDEV_PRE_TYPE_CHANGE, 2500 NETDEV_POST_TYPE_CHANGE, 2501 NETDEV_POST_INIT, 2502 NETDEV_RELEASE, 2503 NETDEV_NOTIFY_PEERS, 2504 NETDEV_JOIN, 2505 NETDEV_CHANGEUPPER, 2506 NETDEV_RESEND_IGMP, 2507 NETDEV_PRECHANGEMTU, /* notify before mtu change happened */ 2508 NETDEV_CHANGEINFODATA, 2509 NETDEV_BONDING_INFO, 2510 NETDEV_PRECHANGEUPPER, 2511 NETDEV_CHANGELOWERSTATE, 2512 NETDEV_UDP_TUNNEL_PUSH_INFO, 2513 NETDEV_UDP_TUNNEL_DROP_INFO, 2514 NETDEV_CHANGE_TX_QUEUE_LEN, 2515 NETDEV_CVLAN_FILTER_PUSH_INFO, 2516 NETDEV_CVLAN_FILTER_DROP_INFO, 2517 NETDEV_SVLAN_FILTER_PUSH_INFO, 2518 NETDEV_SVLAN_FILTER_DROP_INFO, 2519 }; 2520 const char *netdev_cmd_to_name(enum netdev_cmd cmd); 2521 2522 int register_netdevice_notifier(struct notifier_block *nb); 2523 int unregister_netdevice_notifier(struct notifier_block *nb); 2524 int register_netdevice_notifier_net(struct net *net, struct notifier_block *nb); 2525 int unregister_netdevice_notifier_net(struct net *net, 2526 struct notifier_block *nb); 2527 2528 struct netdev_notifier_info { 2529 struct net_device *dev; 2530 struct netlink_ext_ack *extack; 2531 }; 2532 2533 struct netdev_notifier_info_ext { 2534 struct netdev_notifier_info info; /* must be first */ 2535 union { 2536 u32 mtu; 2537 } ext; 2538 }; 2539 2540 struct netdev_notifier_change_info { 2541 struct netdev_notifier_info info; /* must be first */ 2542 unsigned int flags_changed; 2543 }; 2544 2545 struct netdev_notifier_changeupper_info { 2546 struct netdev_notifier_info info; /* must be first */ 2547 struct net_device *upper_dev; /* new upper dev */ 2548 bool master; /* is upper dev master */ 2549 bool linking; /* is the notification for link or unlink */ 2550 void *upper_info; /* upper dev info */ 2551 }; 2552 2553 struct netdev_notifier_changelowerstate_info { 2554 struct netdev_notifier_info info; /* must be first */ 2555 void *lower_state_info; /* is lower dev state */ 2556 }; 2557 2558 struct netdev_notifier_pre_changeaddr_info { 2559 struct netdev_notifier_info info; /* must be first */ 2560 const unsigned char *dev_addr; 2561 }; 2562 2563 static inline void netdev_notifier_info_init(struct netdev_notifier_info *info, 2564 struct net_device *dev) 2565 { 2566 info->dev = dev; 2567 info->extack = NULL; 2568 } 2569 2570 static inline struct net_device * 2571 netdev_notifier_info_to_dev(const struct netdev_notifier_info *info) 2572 { 2573 return info->dev; 2574 } 2575 2576 static inline struct netlink_ext_ack * 2577 netdev_notifier_info_to_extack(const struct netdev_notifier_info *info) 2578 { 2579 return info->extack; 2580 } 2581 2582 int call_netdevice_notifiers(unsigned long val, struct net_device *dev); 2583 2584 2585 extern rwlock_t dev_base_lock; /* Device list lock */ 2586 2587 #define for_each_netdev(net, d) \ 2588 list_for_each_entry(d, &(net)->dev_base_head, dev_list) 2589 #define for_each_netdev_reverse(net, d) \ 2590 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list) 2591 #define for_each_netdev_rcu(net, d) \ 2592 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list) 2593 #define for_each_netdev_safe(net, d, n) \ 2594 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list) 2595 #define for_each_netdev_continue(net, d) \ 2596 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list) 2597 #define for_each_netdev_continue_reverse(net, d) \ 2598 list_for_each_entry_continue_reverse(d, &(net)->dev_base_head, \ 2599 dev_list) 2600 #define for_each_netdev_continue_rcu(net, d) \ 2601 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list) 2602 #define for_each_netdev_in_bond_rcu(bond, slave) \ 2603 for_each_netdev_rcu(&init_net, slave) \ 2604 if (netdev_master_upper_dev_get_rcu(slave) == (bond)) 2605 #define net_device_entry(lh) list_entry(lh, struct net_device, dev_list) 2606 2607 static inline struct net_device *next_net_device(struct net_device *dev) 2608 { 2609 struct list_head *lh; 2610 struct net *net; 2611 2612 net = dev_net(dev); 2613 lh = dev->dev_list.next; 2614 return lh == &net->dev_base_head ? NULL : net_device_entry(lh); 2615 } 2616 2617 static inline struct net_device *next_net_device_rcu(struct net_device *dev) 2618 { 2619 struct list_head *lh; 2620 struct net *net; 2621 2622 net = dev_net(dev); 2623 lh = rcu_dereference(list_next_rcu(&dev->dev_list)); 2624 return lh == &net->dev_base_head ? NULL : net_device_entry(lh); 2625 } 2626 2627 static inline struct net_device *first_net_device(struct net *net) 2628 { 2629 return list_empty(&net->dev_base_head) ? NULL : 2630 net_device_entry(net->dev_base_head.next); 2631 } 2632 2633 static inline struct net_device *first_net_device_rcu(struct net *net) 2634 { 2635 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head)); 2636 2637 return lh == &net->dev_base_head ? NULL : net_device_entry(lh); 2638 } 2639 2640 int netdev_boot_setup_check(struct net_device *dev); 2641 unsigned long netdev_boot_base(const char *prefix, int unit); 2642 struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type, 2643 const char *hwaddr); 2644 struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type); 2645 struct net_device *__dev_getfirstbyhwtype(struct net *net, unsigned short type); 2646 void dev_add_pack(struct packet_type *pt); 2647 void dev_remove_pack(struct packet_type *pt); 2648 void __dev_remove_pack(struct packet_type *pt); 2649 void dev_add_offload(struct packet_offload *po); 2650 void dev_remove_offload(struct packet_offload *po); 2651 2652 int dev_get_iflink(const struct net_device *dev); 2653 int dev_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb); 2654 struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags, 2655 unsigned short mask); 2656 struct net_device *dev_get_by_name(struct net *net, const char *name); 2657 struct net_device *dev_get_by_name_rcu(struct net *net, const char *name); 2658 struct net_device *__dev_get_by_name(struct net *net, const char *name); 2659 int dev_alloc_name(struct net_device *dev, const char *name); 2660 int dev_open(struct net_device *dev, struct netlink_ext_ack *extack); 2661 void dev_close(struct net_device *dev); 2662 void dev_close_many(struct list_head *head, bool unlink); 2663 void dev_disable_lro(struct net_device *dev); 2664 int dev_loopback_xmit(struct net *net, struct sock *sk, struct sk_buff *newskb); 2665 u16 dev_pick_tx_zero(struct net_device *dev, struct sk_buff *skb, 2666 struct net_device *sb_dev); 2667 u16 dev_pick_tx_cpu_id(struct net_device *dev, struct sk_buff *skb, 2668 struct net_device *sb_dev); 2669 int dev_queue_xmit(struct sk_buff *skb); 2670 int dev_queue_xmit_accel(struct sk_buff *skb, struct net_device *sb_dev); 2671 int dev_direct_xmit(struct sk_buff *skb, u16 queue_id); 2672 int register_netdevice(struct net_device *dev); 2673 void unregister_netdevice_queue(struct net_device *dev, struct list_head *head); 2674 void unregister_netdevice_many(struct list_head *head); 2675 static inline void unregister_netdevice(struct net_device *dev) 2676 { 2677 unregister_netdevice_queue(dev, NULL); 2678 } 2679 2680 int netdev_refcnt_read(const struct net_device *dev); 2681 void free_netdev(struct net_device *dev); 2682 void netdev_freemem(struct net_device *dev); 2683 void synchronize_net(void); 2684 int init_dummy_netdev(struct net_device *dev); 2685 2686 struct net_device *dev_get_by_index(struct net *net, int ifindex); 2687 struct net_device *__dev_get_by_index(struct net *net, int ifindex); 2688 struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex); 2689 struct net_device *dev_get_by_napi_id(unsigned int napi_id); 2690 int netdev_get_name(struct net *net, char *name, int ifindex); 2691 int dev_restart(struct net_device *dev); 2692 int skb_gro_receive(struct sk_buff *p, struct sk_buff *skb); 2693 2694 static inline unsigned int skb_gro_offset(const struct sk_buff *skb) 2695 { 2696 return NAPI_GRO_CB(skb)->data_offset; 2697 } 2698 2699 static inline unsigned int skb_gro_len(const struct sk_buff *skb) 2700 { 2701 return skb->len - NAPI_GRO_CB(skb)->data_offset; 2702 } 2703 2704 static inline void skb_gro_pull(struct sk_buff *skb, unsigned int len) 2705 { 2706 NAPI_GRO_CB(skb)->data_offset += len; 2707 } 2708 2709 static inline void *skb_gro_header_fast(struct sk_buff *skb, 2710 unsigned int offset) 2711 { 2712 return NAPI_GRO_CB(skb)->frag0 + offset; 2713 } 2714 2715 static inline int skb_gro_header_hard(struct sk_buff *skb, unsigned int hlen) 2716 { 2717 return NAPI_GRO_CB(skb)->frag0_len < hlen; 2718 } 2719 2720 static inline void skb_gro_frag0_invalidate(struct sk_buff *skb) 2721 { 2722 NAPI_GRO_CB(skb)->frag0 = NULL; 2723 NAPI_GRO_CB(skb)->frag0_len = 0; 2724 } 2725 2726 static inline void *skb_gro_header_slow(struct sk_buff *skb, unsigned int hlen, 2727 unsigned int offset) 2728 { 2729 if (!pskb_may_pull(skb, hlen)) 2730 return NULL; 2731 2732 skb_gro_frag0_invalidate(skb); 2733 return skb->data + offset; 2734 } 2735 2736 static inline void *skb_gro_network_header(struct sk_buff *skb) 2737 { 2738 return (NAPI_GRO_CB(skb)->frag0 ?: skb->data) + 2739 skb_network_offset(skb); 2740 } 2741 2742 static inline void skb_gro_postpull_rcsum(struct sk_buff *skb, 2743 const void *start, unsigned int len) 2744 { 2745 if (NAPI_GRO_CB(skb)->csum_valid) 2746 NAPI_GRO_CB(skb)->csum = csum_sub(NAPI_GRO_CB(skb)->csum, 2747 csum_partial(start, len, 0)); 2748 } 2749 2750 /* GRO checksum functions. These are logical equivalents of the normal 2751 * checksum functions (in skbuff.h) except that they operate on the GRO 2752 * offsets and fields in sk_buff. 2753 */ 2754 2755 __sum16 __skb_gro_checksum_complete(struct sk_buff *skb); 2756 2757 static inline bool skb_at_gro_remcsum_start(struct sk_buff *skb) 2758 { 2759 return (NAPI_GRO_CB(skb)->gro_remcsum_start == skb_gro_offset(skb)); 2760 } 2761 2762 static inline bool __skb_gro_checksum_validate_needed(struct sk_buff *skb, 2763 bool zero_okay, 2764 __sum16 check) 2765 { 2766 return ((skb->ip_summed != CHECKSUM_PARTIAL || 2767 skb_checksum_start_offset(skb) < 2768 skb_gro_offset(skb)) && 2769 !skb_at_gro_remcsum_start(skb) && 2770 NAPI_GRO_CB(skb)->csum_cnt == 0 && 2771 (!zero_okay || check)); 2772 } 2773 2774 static inline __sum16 __skb_gro_checksum_validate_complete(struct sk_buff *skb, 2775 __wsum psum) 2776 { 2777 if (NAPI_GRO_CB(skb)->csum_valid && 2778 !csum_fold(csum_add(psum, NAPI_GRO_CB(skb)->csum))) 2779 return 0; 2780 2781 NAPI_GRO_CB(skb)->csum = psum; 2782 2783 return __skb_gro_checksum_complete(skb); 2784 } 2785 2786 static inline void skb_gro_incr_csum_unnecessary(struct sk_buff *skb) 2787 { 2788 if (NAPI_GRO_CB(skb)->csum_cnt > 0) { 2789 /* Consume a checksum from CHECKSUM_UNNECESSARY */ 2790 NAPI_GRO_CB(skb)->csum_cnt--; 2791 } else { 2792 /* Update skb for CHECKSUM_UNNECESSARY and csum_level when we 2793 * verified a new top level checksum or an encapsulated one 2794 * during GRO. This saves work if we fallback to normal path. 2795 */ 2796 __skb_incr_checksum_unnecessary(skb); 2797 } 2798 } 2799 2800 #define __skb_gro_checksum_validate(skb, proto, zero_okay, check, \ 2801 compute_pseudo) \ 2802 ({ \ 2803 __sum16 __ret = 0; \ 2804 if (__skb_gro_checksum_validate_needed(skb, zero_okay, check)) \ 2805 __ret = __skb_gro_checksum_validate_complete(skb, \ 2806 compute_pseudo(skb, proto)); \ 2807 if (!__ret) \ 2808 skb_gro_incr_csum_unnecessary(skb); \ 2809 __ret; \ 2810 }) 2811 2812 #define skb_gro_checksum_validate(skb, proto, compute_pseudo) \ 2813 __skb_gro_checksum_validate(skb, proto, false, 0, compute_pseudo) 2814 2815 #define skb_gro_checksum_validate_zero_check(skb, proto, check, \ 2816 compute_pseudo) \ 2817 __skb_gro_checksum_validate(skb, proto, true, check, compute_pseudo) 2818 2819 #define skb_gro_checksum_simple_validate(skb) \ 2820 __skb_gro_checksum_validate(skb, 0, false, 0, null_compute_pseudo) 2821 2822 static inline bool __skb_gro_checksum_convert_check(struct sk_buff *skb) 2823 { 2824 return (NAPI_GRO_CB(skb)->csum_cnt == 0 && 2825 !NAPI_GRO_CB(skb)->csum_valid); 2826 } 2827 2828 static inline void __skb_gro_checksum_convert(struct sk_buff *skb, 2829 __wsum pseudo) 2830 { 2831 NAPI_GRO_CB(skb)->csum = ~pseudo; 2832 NAPI_GRO_CB(skb)->csum_valid = 1; 2833 } 2834 2835 #define skb_gro_checksum_try_convert(skb, proto, compute_pseudo) \ 2836 do { \ 2837 if (__skb_gro_checksum_convert_check(skb)) \ 2838 __skb_gro_checksum_convert(skb, \ 2839 compute_pseudo(skb, proto)); \ 2840 } while (0) 2841 2842 struct gro_remcsum { 2843 int offset; 2844 __wsum delta; 2845 }; 2846 2847 static inline void skb_gro_remcsum_init(struct gro_remcsum *grc) 2848 { 2849 grc->offset = 0; 2850 grc->delta = 0; 2851 } 2852 2853 static inline void *skb_gro_remcsum_process(struct sk_buff *skb, void *ptr, 2854 unsigned int off, size_t hdrlen, 2855 int start, int offset, 2856 struct gro_remcsum *grc, 2857 bool nopartial) 2858 { 2859 __wsum delta; 2860 size_t plen = hdrlen + max_t(size_t, offset + sizeof(u16), start); 2861 2862 BUG_ON(!NAPI_GRO_CB(skb)->csum_valid); 2863 2864 if (!nopartial) { 2865 NAPI_GRO_CB(skb)->gro_remcsum_start = off + hdrlen + start; 2866 return ptr; 2867 } 2868 2869 ptr = skb_gro_header_fast(skb, off); 2870 if (skb_gro_header_hard(skb, off + plen)) { 2871 ptr = skb_gro_header_slow(skb, off + plen, off); 2872 if (!ptr) 2873 return NULL; 2874 } 2875 2876 delta = remcsum_adjust(ptr + hdrlen, NAPI_GRO_CB(skb)->csum, 2877 start, offset); 2878 2879 /* Adjust skb->csum since we changed the packet */ 2880 NAPI_GRO_CB(skb)->csum = csum_add(NAPI_GRO_CB(skb)->csum, delta); 2881 2882 grc->offset = off + hdrlen + offset; 2883 grc->delta = delta; 2884 2885 return ptr; 2886 } 2887 2888 static inline void skb_gro_remcsum_cleanup(struct sk_buff *skb, 2889 struct gro_remcsum *grc) 2890 { 2891 void *ptr; 2892 size_t plen = grc->offset + sizeof(u16); 2893 2894 if (!grc->delta) 2895 return; 2896 2897 ptr = skb_gro_header_fast(skb, grc->offset); 2898 if (skb_gro_header_hard(skb, grc->offset + sizeof(u16))) { 2899 ptr = skb_gro_header_slow(skb, plen, grc->offset); 2900 if (!ptr) 2901 return; 2902 } 2903 2904 remcsum_unadjust((__sum16 *)ptr, grc->delta); 2905 } 2906 2907 #ifdef CONFIG_XFRM_OFFLOAD 2908 static inline void skb_gro_flush_final(struct sk_buff *skb, struct sk_buff *pp, int flush) 2909 { 2910 if (PTR_ERR(pp) != -EINPROGRESS) 2911 NAPI_GRO_CB(skb)->flush |= flush; 2912 } 2913 static inline void skb_gro_flush_final_remcsum(struct sk_buff *skb, 2914 struct sk_buff *pp, 2915 int flush, 2916 struct gro_remcsum *grc) 2917 { 2918 if (PTR_ERR(pp) != -EINPROGRESS) { 2919 NAPI_GRO_CB(skb)->flush |= flush; 2920 skb_gro_remcsum_cleanup(skb, grc); 2921 skb->remcsum_offload = 0; 2922 } 2923 } 2924 #else 2925 static inline void skb_gro_flush_final(struct sk_buff *skb, struct sk_buff *pp, int flush) 2926 { 2927 NAPI_GRO_CB(skb)->flush |= flush; 2928 } 2929 static inline void skb_gro_flush_final_remcsum(struct sk_buff *skb, 2930 struct sk_buff *pp, 2931 int flush, 2932 struct gro_remcsum *grc) 2933 { 2934 NAPI_GRO_CB(skb)->flush |= flush; 2935 skb_gro_remcsum_cleanup(skb, grc); 2936 skb->remcsum_offload = 0; 2937 } 2938 #endif 2939 2940 static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev, 2941 unsigned short type, 2942 const void *daddr, const void *saddr, 2943 unsigned int len) 2944 { 2945 if (!dev->header_ops || !dev->header_ops->create) 2946 return 0; 2947 2948 return dev->header_ops->create(skb, dev, type, daddr, saddr, len); 2949 } 2950 2951 static inline int dev_parse_header(const struct sk_buff *skb, 2952 unsigned char *haddr) 2953 { 2954 const struct net_device *dev = skb->dev; 2955 2956 if (!dev->header_ops || !dev->header_ops->parse) 2957 return 0; 2958 return dev->header_ops->parse(skb, haddr); 2959 } 2960 2961 static inline __be16 dev_parse_header_protocol(const struct sk_buff *skb) 2962 { 2963 const struct net_device *dev = skb->dev; 2964 2965 if (!dev->header_ops || !dev->header_ops->parse_protocol) 2966 return 0; 2967 return dev->header_ops->parse_protocol(skb); 2968 } 2969 2970 /* ll_header must have at least hard_header_len allocated */ 2971 static inline bool dev_validate_header(const struct net_device *dev, 2972 char *ll_header, int len) 2973 { 2974 if (likely(len >= dev->hard_header_len)) 2975 return true; 2976 if (len < dev->min_header_len) 2977 return false; 2978 2979 if (capable(CAP_SYS_RAWIO)) { 2980 memset(ll_header + len, 0, dev->hard_header_len - len); 2981 return true; 2982 } 2983 2984 if (dev->header_ops && dev->header_ops->validate) 2985 return dev->header_ops->validate(ll_header, len); 2986 2987 return false; 2988 } 2989 2990 typedef int gifconf_func_t(struct net_device * dev, char __user * bufptr, 2991 int len, int size); 2992 int register_gifconf(unsigned int family, gifconf_func_t *gifconf); 2993 static inline int unregister_gifconf(unsigned int family) 2994 { 2995 return register_gifconf(family, NULL); 2996 } 2997 2998 #ifdef CONFIG_NET_FLOW_LIMIT 2999 #define FLOW_LIMIT_HISTORY (1 << 7) /* must be ^2 and !overflow buckets */ 3000 struct sd_flow_limit { 3001 u64 count; 3002 unsigned int num_buckets; 3003 unsigned int history_head; 3004 u16 history[FLOW_LIMIT_HISTORY]; 3005 u8 buckets[]; 3006 }; 3007 3008 extern int netdev_flow_limit_table_len; 3009 #endif /* CONFIG_NET_FLOW_LIMIT */ 3010 3011 /* 3012 * Incoming packets are placed on per-CPU queues 3013 */ 3014 struct softnet_data { 3015 struct list_head poll_list; 3016 struct sk_buff_head process_queue; 3017 3018 /* stats */ 3019 unsigned int processed; 3020 unsigned int time_squeeze; 3021 unsigned int received_rps; 3022 #ifdef CONFIG_RPS 3023 struct softnet_data *rps_ipi_list; 3024 #endif 3025 #ifdef CONFIG_NET_FLOW_LIMIT 3026 struct sd_flow_limit __rcu *flow_limit; 3027 #endif 3028 struct Qdisc *output_queue; 3029 struct Qdisc **output_queue_tailp; 3030 struct sk_buff *completion_queue; 3031 #ifdef CONFIG_XFRM_OFFLOAD 3032 struct sk_buff_head xfrm_backlog; 3033 #endif 3034 /* written and read only by owning cpu: */ 3035 struct { 3036 u16 recursion; 3037 u8 more; 3038 } xmit; 3039 #ifdef CONFIG_RPS 3040 /* input_queue_head should be written by cpu owning this struct, 3041 * and only read by other cpus. Worth using a cache line. 3042 */ 3043 unsigned int input_queue_head ____cacheline_aligned_in_smp; 3044 3045 /* Elements below can be accessed between CPUs for RPS/RFS */ 3046 call_single_data_t csd ____cacheline_aligned_in_smp; 3047 struct softnet_data *rps_ipi_next; 3048 unsigned int cpu; 3049 unsigned int input_queue_tail; 3050 #endif 3051 unsigned int dropped; 3052 struct sk_buff_head input_pkt_queue; 3053 struct napi_struct backlog; 3054 3055 }; 3056 3057 static inline void input_queue_head_incr(struct softnet_data *sd) 3058 { 3059 #ifdef CONFIG_RPS 3060 sd->input_queue_head++; 3061 #endif 3062 } 3063 3064 static inline void input_queue_tail_incr_save(struct softnet_data *sd, 3065 unsigned int *qtail) 3066 { 3067 #ifdef CONFIG_RPS 3068 *qtail = ++sd->input_queue_tail; 3069 #endif 3070 } 3071 3072 DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data); 3073 3074 static inline int dev_recursion_level(void) 3075 { 3076 return this_cpu_read(softnet_data.xmit.recursion); 3077 } 3078 3079 #define XMIT_RECURSION_LIMIT 10 3080 static inline bool dev_xmit_recursion(void) 3081 { 3082 return unlikely(__this_cpu_read(softnet_data.xmit.recursion) > 3083 XMIT_RECURSION_LIMIT); 3084 } 3085 3086 static inline void dev_xmit_recursion_inc(void) 3087 { 3088 __this_cpu_inc(softnet_data.xmit.recursion); 3089 } 3090 3091 static inline void dev_xmit_recursion_dec(void) 3092 { 3093 __this_cpu_dec(softnet_data.xmit.recursion); 3094 } 3095 3096 void __netif_schedule(struct Qdisc *q); 3097 void netif_schedule_queue(struct netdev_queue *txq); 3098 3099 static inline void netif_tx_schedule_all(struct net_device *dev) 3100 { 3101 unsigned int i; 3102 3103 for (i = 0; i < dev->num_tx_queues; i++) 3104 netif_schedule_queue(netdev_get_tx_queue(dev, i)); 3105 } 3106 3107 static __always_inline void netif_tx_start_queue(struct netdev_queue *dev_queue) 3108 { 3109 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state); 3110 } 3111 3112 /** 3113 * netif_start_queue - allow transmit 3114 * @dev: network device 3115 * 3116 * Allow upper layers to call the device hard_start_xmit routine. 3117 */ 3118 static inline void netif_start_queue(struct net_device *dev) 3119 { 3120 netif_tx_start_queue(netdev_get_tx_queue(dev, 0)); 3121 } 3122 3123 static inline void netif_tx_start_all_queues(struct net_device *dev) 3124 { 3125 unsigned int i; 3126 3127 for (i = 0; i < dev->num_tx_queues; i++) { 3128 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 3129 netif_tx_start_queue(txq); 3130 } 3131 } 3132 3133 void netif_tx_wake_queue(struct netdev_queue *dev_queue); 3134 3135 /** 3136 * netif_wake_queue - restart transmit 3137 * @dev: network device 3138 * 3139 * Allow upper layers to call the device hard_start_xmit routine. 3140 * Used for flow control when transmit resources are available. 3141 */ 3142 static inline void netif_wake_queue(struct net_device *dev) 3143 { 3144 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0)); 3145 } 3146 3147 static inline void netif_tx_wake_all_queues(struct net_device *dev) 3148 { 3149 unsigned int i; 3150 3151 for (i = 0; i < dev->num_tx_queues; i++) { 3152 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 3153 netif_tx_wake_queue(txq); 3154 } 3155 } 3156 3157 static __always_inline void netif_tx_stop_queue(struct netdev_queue *dev_queue) 3158 { 3159 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state); 3160 } 3161 3162 /** 3163 * netif_stop_queue - stop transmitted packets 3164 * @dev: network device 3165 * 3166 * Stop upper layers calling the device hard_start_xmit routine. 3167 * Used for flow control when transmit resources are unavailable. 3168 */ 3169 static inline void netif_stop_queue(struct net_device *dev) 3170 { 3171 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0)); 3172 } 3173 3174 void netif_tx_stop_all_queues(struct net_device *dev); 3175 void netdev_update_lockdep_key(struct net_device *dev); 3176 3177 static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue) 3178 { 3179 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state); 3180 } 3181 3182 /** 3183 * netif_queue_stopped - test if transmit queue is flowblocked 3184 * @dev: network device 3185 * 3186 * Test if transmit queue on device is currently unable to send. 3187 */ 3188 static inline bool netif_queue_stopped(const struct net_device *dev) 3189 { 3190 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0)); 3191 } 3192 3193 static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue) 3194 { 3195 return dev_queue->state & QUEUE_STATE_ANY_XOFF; 3196 } 3197 3198 static inline bool 3199 netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue) 3200 { 3201 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN; 3202 } 3203 3204 static inline bool 3205 netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue) 3206 { 3207 return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN; 3208 } 3209 3210 /** 3211 * netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write 3212 * @dev_queue: pointer to transmit queue 3213 * 3214 * BQL enabled drivers might use this helper in their ndo_start_xmit(), 3215 * to give appropriate hint to the CPU. 3216 */ 3217 static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue) 3218 { 3219 #ifdef CONFIG_BQL 3220 prefetchw(&dev_queue->dql.num_queued); 3221 #endif 3222 } 3223 3224 /** 3225 * netdev_txq_bql_complete_prefetchw - prefetch bql data for write 3226 * @dev_queue: pointer to transmit queue 3227 * 3228 * BQL enabled drivers might use this helper in their TX completion path, 3229 * to give appropriate hint to the CPU. 3230 */ 3231 static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue) 3232 { 3233 #ifdef CONFIG_BQL 3234 prefetchw(&dev_queue->dql.limit); 3235 #endif 3236 } 3237 3238 static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue, 3239 unsigned int bytes) 3240 { 3241 #ifdef CONFIG_BQL 3242 dql_queued(&dev_queue->dql, bytes); 3243 3244 if (likely(dql_avail(&dev_queue->dql) >= 0)) 3245 return; 3246 3247 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state); 3248 3249 /* 3250 * The XOFF flag must be set before checking the dql_avail below, 3251 * because in netdev_tx_completed_queue we update the dql_completed 3252 * before checking the XOFF flag. 3253 */ 3254 smp_mb(); 3255 3256 /* check again in case another CPU has just made room avail */ 3257 if (unlikely(dql_avail(&dev_queue->dql) >= 0)) 3258 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state); 3259 #endif 3260 } 3261 3262 /* Variant of netdev_tx_sent_queue() for drivers that are aware 3263 * that they should not test BQL status themselves. 3264 * We do want to change __QUEUE_STATE_STACK_XOFF only for the last 3265 * skb of a batch. 3266 * Returns true if the doorbell must be used to kick the NIC. 3267 */ 3268 static inline bool __netdev_tx_sent_queue(struct netdev_queue *dev_queue, 3269 unsigned int bytes, 3270 bool xmit_more) 3271 { 3272 if (xmit_more) { 3273 #ifdef CONFIG_BQL 3274 dql_queued(&dev_queue->dql, bytes); 3275 #endif 3276 return netif_tx_queue_stopped(dev_queue); 3277 } 3278 netdev_tx_sent_queue(dev_queue, bytes); 3279 return true; 3280 } 3281 3282 /** 3283 * netdev_sent_queue - report the number of bytes queued to hardware 3284 * @dev: network device 3285 * @bytes: number of bytes queued to the hardware device queue 3286 * 3287 * Report the number of bytes queued for sending/completion to the network 3288 * device hardware queue. @bytes should be a good approximation and should 3289 * exactly match netdev_completed_queue() @bytes 3290 */ 3291 static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes) 3292 { 3293 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes); 3294 } 3295 3296 static inline bool __netdev_sent_queue(struct net_device *dev, 3297 unsigned int bytes, 3298 bool xmit_more) 3299 { 3300 return __netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes, 3301 xmit_more); 3302 } 3303 3304 static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue, 3305 unsigned int pkts, unsigned int bytes) 3306 { 3307 #ifdef CONFIG_BQL 3308 if (unlikely(!bytes)) 3309 return; 3310 3311 dql_completed(&dev_queue->dql, bytes); 3312 3313 /* 3314 * Without the memory barrier there is a small possiblity that 3315 * netdev_tx_sent_queue will miss the update and cause the queue to 3316 * be stopped forever 3317 */ 3318 smp_mb(); 3319 3320 if (unlikely(dql_avail(&dev_queue->dql) < 0)) 3321 return; 3322 3323 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state)) 3324 netif_schedule_queue(dev_queue); 3325 #endif 3326 } 3327 3328 /** 3329 * netdev_completed_queue - report bytes and packets completed by device 3330 * @dev: network device 3331 * @pkts: actual number of packets sent over the medium 3332 * @bytes: actual number of bytes sent over the medium 3333 * 3334 * Report the number of bytes and packets transmitted by the network device 3335 * hardware queue over the physical medium, @bytes must exactly match the 3336 * @bytes amount passed to netdev_sent_queue() 3337 */ 3338 static inline void netdev_completed_queue(struct net_device *dev, 3339 unsigned int pkts, unsigned int bytes) 3340 { 3341 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes); 3342 } 3343 3344 static inline void netdev_tx_reset_queue(struct netdev_queue *q) 3345 { 3346 #ifdef CONFIG_BQL 3347 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state); 3348 dql_reset(&q->dql); 3349 #endif 3350 } 3351 3352 /** 3353 * netdev_reset_queue - reset the packets and bytes count of a network device 3354 * @dev_queue: network device 3355 * 3356 * Reset the bytes and packet count of a network device and clear the 3357 * software flow control OFF bit for this network device 3358 */ 3359 static inline void netdev_reset_queue(struct net_device *dev_queue) 3360 { 3361 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0)); 3362 } 3363 3364 /** 3365 * netdev_cap_txqueue - check if selected tx queue exceeds device queues 3366 * @dev: network device 3367 * @queue_index: given tx queue index 3368 * 3369 * Returns 0 if given tx queue index >= number of device tx queues, 3370 * otherwise returns the originally passed tx queue index. 3371 */ 3372 static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index) 3373 { 3374 if (unlikely(queue_index >= dev->real_num_tx_queues)) { 3375 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n", 3376 dev->name, queue_index, 3377 dev->real_num_tx_queues); 3378 return 0; 3379 } 3380 3381 return queue_index; 3382 } 3383 3384 /** 3385 * netif_running - test if up 3386 * @dev: network device 3387 * 3388 * Test if the device has been brought up. 3389 */ 3390 static inline bool netif_running(const struct net_device *dev) 3391 { 3392 return test_bit(__LINK_STATE_START, &dev->state); 3393 } 3394 3395 /* 3396 * Routines to manage the subqueues on a device. We only need start, 3397 * stop, and a check if it's stopped. All other device management is 3398 * done at the overall netdevice level. 3399 * Also test the device if we're multiqueue. 3400 */ 3401 3402 /** 3403 * netif_start_subqueue - allow sending packets on subqueue 3404 * @dev: network device 3405 * @queue_index: sub queue index 3406 * 3407 * Start individual transmit queue of a device with multiple transmit queues. 3408 */ 3409 static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index) 3410 { 3411 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 3412 3413 netif_tx_start_queue(txq); 3414 } 3415 3416 /** 3417 * netif_stop_subqueue - stop sending packets on subqueue 3418 * @dev: network device 3419 * @queue_index: sub queue index 3420 * 3421 * Stop individual transmit queue of a device with multiple transmit queues. 3422 */ 3423 static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index) 3424 { 3425 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 3426 netif_tx_stop_queue(txq); 3427 } 3428 3429 /** 3430 * netif_subqueue_stopped - test status of subqueue 3431 * @dev: network device 3432 * @queue_index: sub queue index 3433 * 3434 * Check individual transmit queue of a device with multiple transmit queues. 3435 */ 3436 static inline bool __netif_subqueue_stopped(const struct net_device *dev, 3437 u16 queue_index) 3438 { 3439 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 3440 3441 return netif_tx_queue_stopped(txq); 3442 } 3443 3444 static inline bool netif_subqueue_stopped(const struct net_device *dev, 3445 struct sk_buff *skb) 3446 { 3447 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb)); 3448 } 3449 3450 /** 3451 * netif_wake_subqueue - allow sending packets on subqueue 3452 * @dev: network device 3453 * @queue_index: sub queue index 3454 * 3455 * Resume individual transmit queue of a device with multiple transmit queues. 3456 */ 3457 static inline void netif_wake_subqueue(struct net_device *dev, u16 queue_index) 3458 { 3459 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 3460 3461 netif_tx_wake_queue(txq); 3462 } 3463 3464 #ifdef CONFIG_XPS 3465 int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask, 3466 u16 index); 3467 int __netif_set_xps_queue(struct net_device *dev, const unsigned long *mask, 3468 u16 index, bool is_rxqs_map); 3469 3470 /** 3471 * netif_attr_test_mask - Test a CPU or Rx queue set in a mask 3472 * @j: CPU/Rx queue index 3473 * @mask: bitmask of all cpus/rx queues 3474 * @nr_bits: number of bits in the bitmask 3475 * 3476 * Test if a CPU or Rx queue index is set in a mask of all CPU/Rx queues. 3477 */ 3478 static inline bool netif_attr_test_mask(unsigned long j, 3479 const unsigned long *mask, 3480 unsigned int nr_bits) 3481 { 3482 cpu_max_bits_warn(j, nr_bits); 3483 return test_bit(j, mask); 3484 } 3485 3486 /** 3487 * netif_attr_test_online - Test for online CPU/Rx queue 3488 * @j: CPU/Rx queue index 3489 * @online_mask: bitmask for CPUs/Rx queues that are online 3490 * @nr_bits: number of bits in the bitmask 3491 * 3492 * Returns true if a CPU/Rx queue is online. 3493 */ 3494 static inline bool netif_attr_test_online(unsigned long j, 3495 const unsigned long *online_mask, 3496 unsigned int nr_bits) 3497 { 3498 cpu_max_bits_warn(j, nr_bits); 3499 3500 if (online_mask) 3501 return test_bit(j, online_mask); 3502 3503 return (j < nr_bits); 3504 } 3505 3506 /** 3507 * netif_attrmask_next - get the next CPU/Rx queue in a cpu/Rx queues mask 3508 * @n: CPU/Rx queue index 3509 * @srcp: the cpumask/Rx queue mask pointer 3510 * @nr_bits: number of bits in the bitmask 3511 * 3512 * Returns >= nr_bits if no further CPUs/Rx queues set. 3513 */ 3514 static inline unsigned int netif_attrmask_next(int n, const unsigned long *srcp, 3515 unsigned int nr_bits) 3516 { 3517 /* -1 is a legal arg here. */ 3518 if (n != -1) 3519 cpu_max_bits_warn(n, nr_bits); 3520 3521 if (srcp) 3522 return find_next_bit(srcp, nr_bits, n + 1); 3523 3524 return n + 1; 3525 } 3526 3527 /** 3528 * netif_attrmask_next_and - get the next CPU/Rx queue in *src1p & *src2p 3529 * @n: CPU/Rx queue index 3530 * @src1p: the first CPUs/Rx queues mask pointer 3531 * @src2p: the second CPUs/Rx queues mask pointer 3532 * @nr_bits: number of bits in the bitmask 3533 * 3534 * Returns >= nr_bits if no further CPUs/Rx queues set in both. 3535 */ 3536 static inline int netif_attrmask_next_and(int n, const unsigned long *src1p, 3537 const unsigned long *src2p, 3538 unsigned int nr_bits) 3539 { 3540 /* -1 is a legal arg here. */ 3541 if (n != -1) 3542 cpu_max_bits_warn(n, nr_bits); 3543 3544 if (src1p && src2p) 3545 return find_next_and_bit(src1p, src2p, nr_bits, n + 1); 3546 else if (src1p) 3547 return find_next_bit(src1p, nr_bits, n + 1); 3548 else if (src2p) 3549 return find_next_bit(src2p, nr_bits, n + 1); 3550 3551 return n + 1; 3552 } 3553 #else 3554 static inline int netif_set_xps_queue(struct net_device *dev, 3555 const struct cpumask *mask, 3556 u16 index) 3557 { 3558 return 0; 3559 } 3560 3561 static inline int __netif_set_xps_queue(struct net_device *dev, 3562 const unsigned long *mask, 3563 u16 index, bool is_rxqs_map) 3564 { 3565 return 0; 3566 } 3567 #endif 3568 3569 /** 3570 * netif_is_multiqueue - test if device has multiple transmit queues 3571 * @dev: network device 3572 * 3573 * Check if device has multiple transmit queues 3574 */ 3575 static inline bool netif_is_multiqueue(const struct net_device *dev) 3576 { 3577 return dev->num_tx_queues > 1; 3578 } 3579 3580 int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq); 3581 3582 #ifdef CONFIG_SYSFS 3583 int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq); 3584 #else 3585 static inline int netif_set_real_num_rx_queues(struct net_device *dev, 3586 unsigned int rxqs) 3587 { 3588 dev->real_num_rx_queues = rxqs; 3589 return 0; 3590 } 3591 #endif 3592 3593 static inline struct netdev_rx_queue * 3594 __netif_get_rx_queue(struct net_device *dev, unsigned int rxq) 3595 { 3596 return dev->_rx + rxq; 3597 } 3598 3599 #ifdef CONFIG_SYSFS 3600 static inline unsigned int get_netdev_rx_queue_index( 3601 struct netdev_rx_queue *queue) 3602 { 3603 struct net_device *dev = queue->dev; 3604 int index = queue - dev->_rx; 3605 3606 BUG_ON(index >= dev->num_rx_queues); 3607 return index; 3608 } 3609 #endif 3610 3611 #define DEFAULT_MAX_NUM_RSS_QUEUES (8) 3612 int netif_get_num_default_rss_queues(void); 3613 3614 enum skb_free_reason { 3615 SKB_REASON_CONSUMED, 3616 SKB_REASON_DROPPED, 3617 }; 3618 3619 void __dev_kfree_skb_irq(struct sk_buff *skb, enum skb_free_reason reason); 3620 void __dev_kfree_skb_any(struct sk_buff *skb, enum skb_free_reason reason); 3621 3622 /* 3623 * It is not allowed to call kfree_skb() or consume_skb() from hardware 3624 * interrupt context or with hardware interrupts being disabled. 3625 * (in_irq() || irqs_disabled()) 3626 * 3627 * We provide four helpers that can be used in following contexts : 3628 * 3629 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context, 3630 * replacing kfree_skb(skb) 3631 * 3632 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context. 3633 * Typically used in place of consume_skb(skb) in TX completion path 3634 * 3635 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context, 3636 * replacing kfree_skb(skb) 3637 * 3638 * dev_consume_skb_any(skb) when caller doesn't know its current irq context, 3639 * and consumed a packet. Used in place of consume_skb(skb) 3640 */ 3641 static inline void dev_kfree_skb_irq(struct sk_buff *skb) 3642 { 3643 __dev_kfree_skb_irq(skb, SKB_REASON_DROPPED); 3644 } 3645 3646 static inline void dev_consume_skb_irq(struct sk_buff *skb) 3647 { 3648 __dev_kfree_skb_irq(skb, SKB_REASON_CONSUMED); 3649 } 3650 3651 static inline void dev_kfree_skb_any(struct sk_buff *skb) 3652 { 3653 __dev_kfree_skb_any(skb, SKB_REASON_DROPPED); 3654 } 3655 3656 static inline void dev_consume_skb_any(struct sk_buff *skb) 3657 { 3658 __dev_kfree_skb_any(skb, SKB_REASON_CONSUMED); 3659 } 3660 3661 void generic_xdp_tx(struct sk_buff *skb, struct bpf_prog *xdp_prog); 3662 int do_xdp_generic(struct bpf_prog *xdp_prog, struct sk_buff *skb); 3663 int netif_rx(struct sk_buff *skb); 3664 int netif_rx_ni(struct sk_buff *skb); 3665 int netif_receive_skb(struct sk_buff *skb); 3666 int netif_receive_skb_core(struct sk_buff *skb); 3667 void netif_receive_skb_list(struct list_head *head); 3668 gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb); 3669 void napi_gro_flush(struct napi_struct *napi, bool flush_old); 3670 struct sk_buff *napi_get_frags(struct napi_struct *napi); 3671 gro_result_t napi_gro_frags(struct napi_struct *napi); 3672 struct packet_offload *gro_find_receive_by_type(__be16 type); 3673 struct packet_offload *gro_find_complete_by_type(__be16 type); 3674 3675 static inline void napi_free_frags(struct napi_struct *napi) 3676 { 3677 kfree_skb(napi->skb); 3678 napi->skb = NULL; 3679 } 3680 3681 bool netdev_is_rx_handler_busy(struct net_device *dev); 3682 int netdev_rx_handler_register(struct net_device *dev, 3683 rx_handler_func_t *rx_handler, 3684 void *rx_handler_data); 3685 void netdev_rx_handler_unregister(struct net_device *dev); 3686 3687 bool dev_valid_name(const char *name); 3688 int dev_ioctl(struct net *net, unsigned int cmd, struct ifreq *ifr, 3689 bool *need_copyout); 3690 int dev_ifconf(struct net *net, struct ifconf *, int); 3691 int dev_ethtool(struct net *net, struct ifreq *); 3692 unsigned int dev_get_flags(const struct net_device *); 3693 int __dev_change_flags(struct net_device *dev, unsigned int flags, 3694 struct netlink_ext_ack *extack); 3695 int dev_change_flags(struct net_device *dev, unsigned int flags, 3696 struct netlink_ext_ack *extack); 3697 void __dev_notify_flags(struct net_device *, unsigned int old_flags, 3698 unsigned int gchanges); 3699 int dev_change_name(struct net_device *, const char *); 3700 int dev_set_alias(struct net_device *, const char *, size_t); 3701 int dev_get_alias(const struct net_device *, char *, size_t); 3702 int dev_change_net_namespace(struct net_device *, struct net *, const char *); 3703 int __dev_set_mtu(struct net_device *, int); 3704 int dev_set_mtu_ext(struct net_device *dev, int mtu, 3705 struct netlink_ext_ack *extack); 3706 int dev_set_mtu(struct net_device *, int); 3707 int dev_change_tx_queue_len(struct net_device *, unsigned long); 3708 void dev_set_group(struct net_device *, int); 3709 int dev_pre_changeaddr_notify(struct net_device *dev, const char *addr, 3710 struct netlink_ext_ack *extack); 3711 int dev_set_mac_address(struct net_device *dev, struct sockaddr *sa, 3712 struct netlink_ext_ack *extack); 3713 int dev_change_carrier(struct net_device *, bool new_carrier); 3714 int dev_get_phys_port_id(struct net_device *dev, 3715 struct netdev_phys_item_id *ppid); 3716 int dev_get_phys_port_name(struct net_device *dev, 3717 char *name, size_t len); 3718 int dev_get_port_parent_id(struct net_device *dev, 3719 struct netdev_phys_item_id *ppid, bool recurse); 3720 bool netdev_port_same_parent_id(struct net_device *a, struct net_device *b); 3721 int dev_change_proto_down(struct net_device *dev, bool proto_down); 3722 int dev_change_proto_down_generic(struct net_device *dev, bool proto_down); 3723 struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev, bool *again); 3724 struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev, 3725 struct netdev_queue *txq, int *ret); 3726 3727 typedef int (*bpf_op_t)(struct net_device *dev, struct netdev_bpf *bpf); 3728 int dev_change_xdp_fd(struct net_device *dev, struct netlink_ext_ack *extack, 3729 int fd, u32 flags); 3730 u32 __dev_xdp_query(struct net_device *dev, bpf_op_t xdp_op, 3731 enum bpf_netdev_command cmd); 3732 int xdp_umem_query(struct net_device *dev, u16 queue_id); 3733 3734 int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb); 3735 int dev_forward_skb(struct net_device *dev, struct sk_buff *skb); 3736 bool is_skb_forwardable(const struct net_device *dev, 3737 const struct sk_buff *skb); 3738 3739 static __always_inline int ____dev_forward_skb(struct net_device *dev, 3740 struct sk_buff *skb) 3741 { 3742 if (skb_orphan_frags(skb, GFP_ATOMIC) || 3743 unlikely(!is_skb_forwardable(dev, skb))) { 3744 atomic_long_inc(&dev->rx_dropped); 3745 kfree_skb(skb); 3746 return NET_RX_DROP; 3747 } 3748 3749 skb_scrub_packet(skb, true); 3750 skb->priority = 0; 3751 return 0; 3752 } 3753 3754 bool dev_nit_active(struct net_device *dev); 3755 void dev_queue_xmit_nit(struct sk_buff *skb, struct net_device *dev); 3756 3757 extern int netdev_budget; 3758 extern unsigned int netdev_budget_usecs; 3759 3760 /* Called by rtnetlink.c:rtnl_unlock() */ 3761 void netdev_run_todo(void); 3762 3763 /** 3764 * dev_put - release reference to device 3765 * @dev: network device 3766 * 3767 * Release reference to device to allow it to be freed. 3768 */ 3769 static inline void dev_put(struct net_device *dev) 3770 { 3771 this_cpu_dec(*dev->pcpu_refcnt); 3772 } 3773 3774 /** 3775 * dev_hold - get reference to device 3776 * @dev: network device 3777 * 3778 * Hold reference to device to keep it from being freed. 3779 */ 3780 static inline void dev_hold(struct net_device *dev) 3781 { 3782 this_cpu_inc(*dev->pcpu_refcnt); 3783 } 3784 3785 /* Carrier loss detection, dial on demand. The functions netif_carrier_on 3786 * and _off may be called from IRQ context, but it is caller 3787 * who is responsible for serialization of these calls. 3788 * 3789 * The name carrier is inappropriate, these functions should really be 3790 * called netif_lowerlayer_*() because they represent the state of any 3791 * kind of lower layer not just hardware media. 3792 */ 3793 3794 void linkwatch_init_dev(struct net_device *dev); 3795 void linkwatch_fire_event(struct net_device *dev); 3796 void linkwatch_forget_dev(struct net_device *dev); 3797 3798 /** 3799 * netif_carrier_ok - test if carrier present 3800 * @dev: network device 3801 * 3802 * Check if carrier is present on device 3803 */ 3804 static inline bool netif_carrier_ok(const struct net_device *dev) 3805 { 3806 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state); 3807 } 3808 3809 unsigned long dev_trans_start(struct net_device *dev); 3810 3811 void __netdev_watchdog_up(struct net_device *dev); 3812 3813 void netif_carrier_on(struct net_device *dev); 3814 3815 void netif_carrier_off(struct net_device *dev); 3816 3817 /** 3818 * netif_dormant_on - mark device as dormant. 3819 * @dev: network device 3820 * 3821 * Mark device as dormant (as per RFC2863). 3822 * 3823 * The dormant state indicates that the relevant interface is not 3824 * actually in a condition to pass packets (i.e., it is not 'up') but is 3825 * in a "pending" state, waiting for some external event. For "on- 3826 * demand" interfaces, this new state identifies the situation where the 3827 * interface is waiting for events to place it in the up state. 3828 */ 3829 static inline void netif_dormant_on(struct net_device *dev) 3830 { 3831 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state)) 3832 linkwatch_fire_event(dev); 3833 } 3834 3835 /** 3836 * netif_dormant_off - set device as not dormant. 3837 * @dev: network device 3838 * 3839 * Device is not in dormant state. 3840 */ 3841 static inline void netif_dormant_off(struct net_device *dev) 3842 { 3843 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state)) 3844 linkwatch_fire_event(dev); 3845 } 3846 3847 /** 3848 * netif_dormant - test if device is dormant 3849 * @dev: network device 3850 * 3851 * Check if device is dormant. 3852 */ 3853 static inline bool netif_dormant(const struct net_device *dev) 3854 { 3855 return test_bit(__LINK_STATE_DORMANT, &dev->state); 3856 } 3857 3858 3859 /** 3860 * netif_oper_up - test if device is operational 3861 * @dev: network device 3862 * 3863 * Check if carrier is operational 3864 */ 3865 static inline bool netif_oper_up(const struct net_device *dev) 3866 { 3867 return (dev->operstate == IF_OPER_UP || 3868 dev->operstate == IF_OPER_UNKNOWN /* backward compat */); 3869 } 3870 3871 /** 3872 * netif_device_present - is device available or removed 3873 * @dev: network device 3874 * 3875 * Check if device has not been removed from system. 3876 */ 3877 static inline bool netif_device_present(struct net_device *dev) 3878 { 3879 return test_bit(__LINK_STATE_PRESENT, &dev->state); 3880 } 3881 3882 void netif_device_detach(struct net_device *dev); 3883 3884 void netif_device_attach(struct net_device *dev); 3885 3886 /* 3887 * Network interface message level settings 3888 */ 3889 3890 enum { 3891 NETIF_MSG_DRV = 0x0001, 3892 NETIF_MSG_PROBE = 0x0002, 3893 NETIF_MSG_LINK = 0x0004, 3894 NETIF_MSG_TIMER = 0x0008, 3895 NETIF_MSG_IFDOWN = 0x0010, 3896 NETIF_MSG_IFUP = 0x0020, 3897 NETIF_MSG_RX_ERR = 0x0040, 3898 NETIF_MSG_TX_ERR = 0x0080, 3899 NETIF_MSG_TX_QUEUED = 0x0100, 3900 NETIF_MSG_INTR = 0x0200, 3901 NETIF_MSG_TX_DONE = 0x0400, 3902 NETIF_MSG_RX_STATUS = 0x0800, 3903 NETIF_MSG_PKTDATA = 0x1000, 3904 NETIF_MSG_HW = 0x2000, 3905 NETIF_MSG_WOL = 0x4000, 3906 }; 3907 3908 #define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV) 3909 #define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE) 3910 #define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK) 3911 #define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER) 3912 #define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN) 3913 #define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP) 3914 #define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR) 3915 #define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR) 3916 #define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED) 3917 #define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR) 3918 #define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE) 3919 #define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS) 3920 #define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA) 3921 #define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW) 3922 #define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL) 3923 3924 static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits) 3925 { 3926 /* use default */ 3927 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8)) 3928 return default_msg_enable_bits; 3929 if (debug_value == 0) /* no output */ 3930 return 0; 3931 /* set low N bits */ 3932 return (1U << debug_value) - 1; 3933 } 3934 3935 static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu) 3936 { 3937 spin_lock(&txq->_xmit_lock); 3938 txq->xmit_lock_owner = cpu; 3939 } 3940 3941 static inline bool __netif_tx_acquire(struct netdev_queue *txq) 3942 { 3943 __acquire(&txq->_xmit_lock); 3944 return true; 3945 } 3946 3947 static inline void __netif_tx_release(struct netdev_queue *txq) 3948 { 3949 __release(&txq->_xmit_lock); 3950 } 3951 3952 static inline void __netif_tx_lock_bh(struct netdev_queue *txq) 3953 { 3954 spin_lock_bh(&txq->_xmit_lock); 3955 txq->xmit_lock_owner = smp_processor_id(); 3956 } 3957 3958 static inline bool __netif_tx_trylock(struct netdev_queue *txq) 3959 { 3960 bool ok = spin_trylock(&txq->_xmit_lock); 3961 if (likely(ok)) 3962 txq->xmit_lock_owner = smp_processor_id(); 3963 return ok; 3964 } 3965 3966 static inline void __netif_tx_unlock(struct netdev_queue *txq) 3967 { 3968 txq->xmit_lock_owner = -1; 3969 spin_unlock(&txq->_xmit_lock); 3970 } 3971 3972 static inline void __netif_tx_unlock_bh(struct netdev_queue *txq) 3973 { 3974 txq->xmit_lock_owner = -1; 3975 spin_unlock_bh(&txq->_xmit_lock); 3976 } 3977 3978 static inline void txq_trans_update(struct netdev_queue *txq) 3979 { 3980 if (txq->xmit_lock_owner != -1) 3981 txq->trans_start = jiffies; 3982 } 3983 3984 /* legacy drivers only, netdev_start_xmit() sets txq->trans_start */ 3985 static inline void netif_trans_update(struct net_device *dev) 3986 { 3987 struct netdev_queue *txq = netdev_get_tx_queue(dev, 0); 3988 3989 if (txq->trans_start != jiffies) 3990 txq->trans_start = jiffies; 3991 } 3992 3993 /** 3994 * netif_tx_lock - grab network device transmit lock 3995 * @dev: network device 3996 * 3997 * Get network device transmit lock 3998 */ 3999 static inline void netif_tx_lock(struct net_device *dev) 4000 { 4001 unsigned int i; 4002 int cpu; 4003 4004 spin_lock(&dev->tx_global_lock); 4005 cpu = smp_processor_id(); 4006 for (i = 0; i < dev->num_tx_queues; i++) { 4007 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 4008 4009 /* We are the only thread of execution doing a 4010 * freeze, but we have to grab the _xmit_lock in 4011 * order to synchronize with threads which are in 4012 * the ->hard_start_xmit() handler and already 4013 * checked the frozen bit. 4014 */ 4015 __netif_tx_lock(txq, cpu); 4016 set_bit(__QUEUE_STATE_FROZEN, &txq->state); 4017 __netif_tx_unlock(txq); 4018 } 4019 } 4020 4021 static inline void netif_tx_lock_bh(struct net_device *dev) 4022 { 4023 local_bh_disable(); 4024 netif_tx_lock(dev); 4025 } 4026 4027 static inline void netif_tx_unlock(struct net_device *dev) 4028 { 4029 unsigned int i; 4030 4031 for (i = 0; i < dev->num_tx_queues; i++) { 4032 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 4033 4034 /* No need to grab the _xmit_lock here. If the 4035 * queue is not stopped for another reason, we 4036 * force a schedule. 4037 */ 4038 clear_bit(__QUEUE_STATE_FROZEN, &txq->state); 4039 netif_schedule_queue(txq); 4040 } 4041 spin_unlock(&dev->tx_global_lock); 4042 } 4043 4044 static inline void netif_tx_unlock_bh(struct net_device *dev) 4045 { 4046 netif_tx_unlock(dev); 4047 local_bh_enable(); 4048 } 4049 4050 #define HARD_TX_LOCK(dev, txq, cpu) { \ 4051 if ((dev->features & NETIF_F_LLTX) == 0) { \ 4052 __netif_tx_lock(txq, cpu); \ 4053 } else { \ 4054 __netif_tx_acquire(txq); \ 4055 } \ 4056 } 4057 4058 #define HARD_TX_TRYLOCK(dev, txq) \ 4059 (((dev->features & NETIF_F_LLTX) == 0) ? \ 4060 __netif_tx_trylock(txq) : \ 4061 __netif_tx_acquire(txq)) 4062 4063 #define HARD_TX_UNLOCK(dev, txq) { \ 4064 if ((dev->features & NETIF_F_LLTX) == 0) { \ 4065 __netif_tx_unlock(txq); \ 4066 } else { \ 4067 __netif_tx_release(txq); \ 4068 } \ 4069 } 4070 4071 static inline void netif_tx_disable(struct net_device *dev) 4072 { 4073 unsigned int i; 4074 int cpu; 4075 4076 local_bh_disable(); 4077 cpu = smp_processor_id(); 4078 for (i = 0; i < dev->num_tx_queues; i++) { 4079 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 4080 4081 __netif_tx_lock(txq, cpu); 4082 netif_tx_stop_queue(txq); 4083 __netif_tx_unlock(txq); 4084 } 4085 local_bh_enable(); 4086 } 4087 4088 static inline void netif_addr_lock(struct net_device *dev) 4089 { 4090 spin_lock(&dev->addr_list_lock); 4091 } 4092 4093 static inline void netif_addr_lock_bh(struct net_device *dev) 4094 { 4095 spin_lock_bh(&dev->addr_list_lock); 4096 } 4097 4098 static inline void netif_addr_unlock(struct net_device *dev) 4099 { 4100 spin_unlock(&dev->addr_list_lock); 4101 } 4102 4103 static inline void netif_addr_unlock_bh(struct net_device *dev) 4104 { 4105 spin_unlock_bh(&dev->addr_list_lock); 4106 } 4107 4108 /* 4109 * dev_addrs walker. Should be used only for read access. Call with 4110 * rcu_read_lock held. 4111 */ 4112 #define for_each_dev_addr(dev, ha) \ 4113 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list) 4114 4115 /* These functions live elsewhere (drivers/net/net_init.c, but related) */ 4116 4117 void ether_setup(struct net_device *dev); 4118 4119 /* Support for loadable net-drivers */ 4120 struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name, 4121 unsigned char name_assign_type, 4122 void (*setup)(struct net_device *), 4123 unsigned int txqs, unsigned int rxqs); 4124 #define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \ 4125 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1) 4126 4127 #define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \ 4128 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \ 4129 count) 4130 4131 int register_netdev(struct net_device *dev); 4132 void unregister_netdev(struct net_device *dev); 4133 4134 /* General hardware address lists handling functions */ 4135 int __hw_addr_sync(struct netdev_hw_addr_list *to_list, 4136 struct netdev_hw_addr_list *from_list, int addr_len); 4137 void __hw_addr_unsync(struct netdev_hw_addr_list *to_list, 4138 struct netdev_hw_addr_list *from_list, int addr_len); 4139 int __hw_addr_sync_dev(struct netdev_hw_addr_list *list, 4140 struct net_device *dev, 4141 int (*sync)(struct net_device *, const unsigned char *), 4142 int (*unsync)(struct net_device *, 4143 const unsigned char *)); 4144 int __hw_addr_ref_sync_dev(struct netdev_hw_addr_list *list, 4145 struct net_device *dev, 4146 int (*sync)(struct net_device *, 4147 const unsigned char *, int), 4148 int (*unsync)(struct net_device *, 4149 const unsigned char *, int)); 4150 void __hw_addr_ref_unsync_dev(struct netdev_hw_addr_list *list, 4151 struct net_device *dev, 4152 int (*unsync)(struct net_device *, 4153 const unsigned char *, int)); 4154 void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list, 4155 struct net_device *dev, 4156 int (*unsync)(struct net_device *, 4157 const unsigned char *)); 4158 void __hw_addr_init(struct netdev_hw_addr_list *list); 4159 4160 /* Functions used for device addresses handling */ 4161 int dev_addr_add(struct net_device *dev, const unsigned char *addr, 4162 unsigned char addr_type); 4163 int dev_addr_del(struct net_device *dev, const unsigned char *addr, 4164 unsigned char addr_type); 4165 void dev_addr_flush(struct net_device *dev); 4166 int dev_addr_init(struct net_device *dev); 4167 4168 /* Functions used for unicast addresses handling */ 4169 int dev_uc_add(struct net_device *dev, const unsigned char *addr); 4170 int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr); 4171 int dev_uc_del(struct net_device *dev, const unsigned char *addr); 4172 int dev_uc_sync(struct net_device *to, struct net_device *from); 4173 int dev_uc_sync_multiple(struct net_device *to, struct net_device *from); 4174 void dev_uc_unsync(struct net_device *to, struct net_device *from); 4175 void dev_uc_flush(struct net_device *dev); 4176 void dev_uc_init(struct net_device *dev); 4177 4178 /** 4179 * __dev_uc_sync - Synchonize device's unicast list 4180 * @dev: device to sync 4181 * @sync: function to call if address should be added 4182 * @unsync: function to call if address should be removed 4183 * 4184 * Add newly added addresses to the interface, and release 4185 * addresses that have been deleted. 4186 */ 4187 static inline int __dev_uc_sync(struct net_device *dev, 4188 int (*sync)(struct net_device *, 4189 const unsigned char *), 4190 int (*unsync)(struct net_device *, 4191 const unsigned char *)) 4192 { 4193 return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync); 4194 } 4195 4196 /** 4197 * __dev_uc_unsync - Remove synchronized addresses from device 4198 * @dev: device to sync 4199 * @unsync: function to call if address should be removed 4200 * 4201 * Remove all addresses that were added to the device by dev_uc_sync(). 4202 */ 4203 static inline void __dev_uc_unsync(struct net_device *dev, 4204 int (*unsync)(struct net_device *, 4205 const unsigned char *)) 4206 { 4207 __hw_addr_unsync_dev(&dev->uc, dev, unsync); 4208 } 4209 4210 /* Functions used for multicast addresses handling */ 4211 int dev_mc_add(struct net_device *dev, const unsigned char *addr); 4212 int dev_mc_add_global(struct net_device *dev, const unsigned char *addr); 4213 int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr); 4214 int dev_mc_del(struct net_device *dev, const unsigned char *addr); 4215 int dev_mc_del_global(struct net_device *dev, const unsigned char *addr); 4216 int dev_mc_sync(struct net_device *to, struct net_device *from); 4217 int dev_mc_sync_multiple(struct net_device *to, struct net_device *from); 4218 void dev_mc_unsync(struct net_device *to, struct net_device *from); 4219 void dev_mc_flush(struct net_device *dev); 4220 void dev_mc_init(struct net_device *dev); 4221 4222 /** 4223 * __dev_mc_sync - Synchonize device's multicast list 4224 * @dev: device to sync 4225 * @sync: function to call if address should be added 4226 * @unsync: function to call if address should be removed 4227 * 4228 * Add newly added addresses to the interface, and release 4229 * addresses that have been deleted. 4230 */ 4231 static inline int __dev_mc_sync(struct net_device *dev, 4232 int (*sync)(struct net_device *, 4233 const unsigned char *), 4234 int (*unsync)(struct net_device *, 4235 const unsigned char *)) 4236 { 4237 return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync); 4238 } 4239 4240 /** 4241 * __dev_mc_unsync - Remove synchronized addresses from device 4242 * @dev: device to sync 4243 * @unsync: function to call if address should be removed 4244 * 4245 * Remove all addresses that were added to the device by dev_mc_sync(). 4246 */ 4247 static inline void __dev_mc_unsync(struct net_device *dev, 4248 int (*unsync)(struct net_device *, 4249 const unsigned char *)) 4250 { 4251 __hw_addr_unsync_dev(&dev->mc, dev, unsync); 4252 } 4253 4254 /* Functions used for secondary unicast and multicast support */ 4255 void dev_set_rx_mode(struct net_device *dev); 4256 void __dev_set_rx_mode(struct net_device *dev); 4257 int dev_set_promiscuity(struct net_device *dev, int inc); 4258 int dev_set_allmulti(struct net_device *dev, int inc); 4259 void netdev_state_change(struct net_device *dev); 4260 void netdev_notify_peers(struct net_device *dev); 4261 void netdev_features_change(struct net_device *dev); 4262 /* Load a device via the kmod */ 4263 void dev_load(struct net *net, const char *name); 4264 struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev, 4265 struct rtnl_link_stats64 *storage); 4266 void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64, 4267 const struct net_device_stats *netdev_stats); 4268 4269 extern int netdev_max_backlog; 4270 extern int netdev_tstamp_prequeue; 4271 extern int weight_p; 4272 extern int dev_weight_rx_bias; 4273 extern int dev_weight_tx_bias; 4274 extern int dev_rx_weight; 4275 extern int dev_tx_weight; 4276 extern int gro_normal_batch; 4277 4278 bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev); 4279 struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev, 4280 struct list_head **iter); 4281 struct net_device *netdev_all_upper_get_next_dev_rcu(struct net_device *dev, 4282 struct list_head **iter); 4283 4284 /* iterate through upper list, must be called under RCU read lock */ 4285 #define netdev_for_each_upper_dev_rcu(dev, updev, iter) \ 4286 for (iter = &(dev)->adj_list.upper, \ 4287 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \ 4288 updev; \ 4289 updev = netdev_upper_get_next_dev_rcu(dev, &(iter))) 4290 4291 int netdev_walk_all_upper_dev_rcu(struct net_device *dev, 4292 int (*fn)(struct net_device *upper_dev, 4293 void *data), 4294 void *data); 4295 4296 bool netdev_has_upper_dev_all_rcu(struct net_device *dev, 4297 struct net_device *upper_dev); 4298 4299 bool netdev_has_any_upper_dev(struct net_device *dev); 4300 4301 void *netdev_lower_get_next_private(struct net_device *dev, 4302 struct list_head **iter); 4303 void *netdev_lower_get_next_private_rcu(struct net_device *dev, 4304 struct list_head **iter); 4305 4306 #define netdev_for_each_lower_private(dev, priv, iter) \ 4307 for (iter = (dev)->adj_list.lower.next, \ 4308 priv = netdev_lower_get_next_private(dev, &(iter)); \ 4309 priv; \ 4310 priv = netdev_lower_get_next_private(dev, &(iter))) 4311 4312 #define netdev_for_each_lower_private_rcu(dev, priv, iter) \ 4313 for (iter = &(dev)->adj_list.lower, \ 4314 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \ 4315 priv; \ 4316 priv = netdev_lower_get_next_private_rcu(dev, &(iter))) 4317 4318 void *netdev_lower_get_next(struct net_device *dev, 4319 struct list_head **iter); 4320 4321 #define netdev_for_each_lower_dev(dev, ldev, iter) \ 4322 for (iter = (dev)->adj_list.lower.next, \ 4323 ldev = netdev_lower_get_next(dev, &(iter)); \ 4324 ldev; \ 4325 ldev = netdev_lower_get_next(dev, &(iter))) 4326 4327 struct net_device *netdev_all_lower_get_next(struct net_device *dev, 4328 struct list_head **iter); 4329 struct net_device *netdev_all_lower_get_next_rcu(struct net_device *dev, 4330 struct list_head **iter); 4331 4332 int netdev_walk_all_lower_dev(struct net_device *dev, 4333 int (*fn)(struct net_device *lower_dev, 4334 void *data), 4335 void *data); 4336 int netdev_walk_all_lower_dev_rcu(struct net_device *dev, 4337 int (*fn)(struct net_device *lower_dev, 4338 void *data), 4339 void *data); 4340 4341 void *netdev_adjacent_get_private(struct list_head *adj_list); 4342 void *netdev_lower_get_first_private_rcu(struct net_device *dev); 4343 struct net_device *netdev_master_upper_dev_get(struct net_device *dev); 4344 struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev); 4345 int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev, 4346 struct netlink_ext_ack *extack); 4347 int netdev_master_upper_dev_link(struct net_device *dev, 4348 struct net_device *upper_dev, 4349 void *upper_priv, void *upper_info, 4350 struct netlink_ext_ack *extack); 4351 void netdev_upper_dev_unlink(struct net_device *dev, 4352 struct net_device *upper_dev); 4353 int netdev_adjacent_change_prepare(struct net_device *old_dev, 4354 struct net_device *new_dev, 4355 struct net_device *dev, 4356 struct netlink_ext_ack *extack); 4357 void netdev_adjacent_change_commit(struct net_device *old_dev, 4358 struct net_device *new_dev, 4359 struct net_device *dev); 4360 void netdev_adjacent_change_abort(struct net_device *old_dev, 4361 struct net_device *new_dev, 4362 struct net_device *dev); 4363 void netdev_adjacent_rename_links(struct net_device *dev, char *oldname); 4364 void *netdev_lower_dev_get_private(struct net_device *dev, 4365 struct net_device *lower_dev); 4366 void netdev_lower_state_changed(struct net_device *lower_dev, 4367 void *lower_state_info); 4368 4369 /* RSS keys are 40 or 52 bytes long */ 4370 #define NETDEV_RSS_KEY_LEN 52 4371 extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN] __read_mostly; 4372 void netdev_rss_key_fill(void *buffer, size_t len); 4373 4374 int skb_checksum_help(struct sk_buff *skb); 4375 int skb_crc32c_csum_help(struct sk_buff *skb); 4376 int skb_csum_hwoffload_help(struct sk_buff *skb, 4377 const netdev_features_t features); 4378 4379 struct sk_buff *__skb_gso_segment(struct sk_buff *skb, 4380 netdev_features_t features, bool tx_path); 4381 struct sk_buff *skb_mac_gso_segment(struct sk_buff *skb, 4382 netdev_features_t features); 4383 4384 struct netdev_bonding_info { 4385 ifslave slave; 4386 ifbond master; 4387 }; 4388 4389 struct netdev_notifier_bonding_info { 4390 struct netdev_notifier_info info; /* must be first */ 4391 struct netdev_bonding_info bonding_info; 4392 }; 4393 4394 void netdev_bonding_info_change(struct net_device *dev, 4395 struct netdev_bonding_info *bonding_info); 4396 4397 #if IS_ENABLED(CONFIG_ETHTOOL_NETLINK) 4398 void ethtool_notify(struct net_device *dev, unsigned int cmd, const void *data); 4399 #else 4400 static inline void ethtool_notify(struct net_device *dev, unsigned int cmd, 4401 const void *data) 4402 { 4403 } 4404 #endif 4405 4406 static inline 4407 struct sk_buff *skb_gso_segment(struct sk_buff *skb, netdev_features_t features) 4408 { 4409 return __skb_gso_segment(skb, features, true); 4410 } 4411 __be16 skb_network_protocol(struct sk_buff *skb, int *depth); 4412 4413 static inline bool can_checksum_protocol(netdev_features_t features, 4414 __be16 protocol) 4415 { 4416 if (protocol == htons(ETH_P_FCOE)) 4417 return !!(features & NETIF_F_FCOE_CRC); 4418 4419 /* Assume this is an IP checksum (not SCTP CRC) */ 4420 4421 if (features & NETIF_F_HW_CSUM) { 4422 /* Can checksum everything */ 4423 return true; 4424 } 4425 4426 switch (protocol) { 4427 case htons(ETH_P_IP): 4428 return !!(features & NETIF_F_IP_CSUM); 4429 case htons(ETH_P_IPV6): 4430 return !!(features & NETIF_F_IPV6_CSUM); 4431 default: 4432 return false; 4433 } 4434 } 4435 4436 #ifdef CONFIG_BUG 4437 void netdev_rx_csum_fault(struct net_device *dev, struct sk_buff *skb); 4438 #else 4439 static inline void netdev_rx_csum_fault(struct net_device *dev, 4440 struct sk_buff *skb) 4441 { 4442 } 4443 #endif 4444 /* rx skb timestamps */ 4445 void net_enable_timestamp(void); 4446 void net_disable_timestamp(void); 4447 4448 #ifdef CONFIG_PROC_FS 4449 int __init dev_proc_init(void); 4450 #else 4451 #define dev_proc_init() 0 4452 #endif 4453 4454 static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops, 4455 struct sk_buff *skb, struct net_device *dev, 4456 bool more) 4457 { 4458 __this_cpu_write(softnet_data.xmit.more, more); 4459 return ops->ndo_start_xmit(skb, dev); 4460 } 4461 4462 static inline bool netdev_xmit_more(void) 4463 { 4464 return __this_cpu_read(softnet_data.xmit.more); 4465 } 4466 4467 static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev, 4468 struct netdev_queue *txq, bool more) 4469 { 4470 const struct net_device_ops *ops = dev->netdev_ops; 4471 netdev_tx_t rc; 4472 4473 rc = __netdev_start_xmit(ops, skb, dev, more); 4474 if (rc == NETDEV_TX_OK) 4475 txq_trans_update(txq); 4476 4477 return rc; 4478 } 4479 4480 int netdev_class_create_file_ns(const struct class_attribute *class_attr, 4481 const void *ns); 4482 void netdev_class_remove_file_ns(const struct class_attribute *class_attr, 4483 const void *ns); 4484 4485 static inline int netdev_class_create_file(const struct class_attribute *class_attr) 4486 { 4487 return netdev_class_create_file_ns(class_attr, NULL); 4488 } 4489 4490 static inline void netdev_class_remove_file(const struct class_attribute *class_attr) 4491 { 4492 netdev_class_remove_file_ns(class_attr, NULL); 4493 } 4494 4495 extern const struct kobj_ns_type_operations net_ns_type_operations; 4496 4497 const char *netdev_drivername(const struct net_device *dev); 4498 4499 void linkwatch_run_queue(void); 4500 4501 static inline netdev_features_t netdev_intersect_features(netdev_features_t f1, 4502 netdev_features_t f2) 4503 { 4504 if ((f1 ^ f2) & NETIF_F_HW_CSUM) { 4505 if (f1 & NETIF_F_HW_CSUM) 4506 f1 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM); 4507 else 4508 f2 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM); 4509 } 4510 4511 return f1 & f2; 4512 } 4513 4514 static inline netdev_features_t netdev_get_wanted_features( 4515 struct net_device *dev) 4516 { 4517 return (dev->features & ~dev->hw_features) | dev->wanted_features; 4518 } 4519 netdev_features_t netdev_increment_features(netdev_features_t all, 4520 netdev_features_t one, netdev_features_t mask); 4521 4522 /* Allow TSO being used on stacked device : 4523 * Performing the GSO segmentation before last device 4524 * is a performance improvement. 4525 */ 4526 static inline netdev_features_t netdev_add_tso_features(netdev_features_t features, 4527 netdev_features_t mask) 4528 { 4529 return netdev_increment_features(features, NETIF_F_ALL_TSO, mask); 4530 } 4531 4532 int __netdev_update_features(struct net_device *dev); 4533 void netdev_update_features(struct net_device *dev); 4534 void netdev_change_features(struct net_device *dev); 4535 4536 void netif_stacked_transfer_operstate(const struct net_device *rootdev, 4537 struct net_device *dev); 4538 4539 netdev_features_t passthru_features_check(struct sk_buff *skb, 4540 struct net_device *dev, 4541 netdev_features_t features); 4542 netdev_features_t netif_skb_features(struct sk_buff *skb); 4543 4544 static inline bool net_gso_ok(netdev_features_t features, int gso_type) 4545 { 4546 netdev_features_t feature = (netdev_features_t)gso_type << NETIF_F_GSO_SHIFT; 4547 4548 /* check flags correspondence */ 4549 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT)); 4550 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT)); 4551 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT)); 4552 BUILD_BUG_ON(SKB_GSO_TCP_FIXEDID != (NETIF_F_TSO_MANGLEID >> NETIF_F_GSO_SHIFT)); 4553 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT)); 4554 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT)); 4555 BUILD_BUG_ON(SKB_GSO_GRE != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT)); 4556 BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT)); 4557 BUILD_BUG_ON(SKB_GSO_IPXIP4 != (NETIF_F_GSO_IPXIP4 >> NETIF_F_GSO_SHIFT)); 4558 BUILD_BUG_ON(SKB_GSO_IPXIP6 != (NETIF_F_GSO_IPXIP6 >> NETIF_F_GSO_SHIFT)); 4559 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT)); 4560 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT)); 4561 BUILD_BUG_ON(SKB_GSO_PARTIAL != (NETIF_F_GSO_PARTIAL >> NETIF_F_GSO_SHIFT)); 4562 BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT)); 4563 BUILD_BUG_ON(SKB_GSO_SCTP != (NETIF_F_GSO_SCTP >> NETIF_F_GSO_SHIFT)); 4564 BUILD_BUG_ON(SKB_GSO_ESP != (NETIF_F_GSO_ESP >> NETIF_F_GSO_SHIFT)); 4565 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_GSO_UDP >> NETIF_F_GSO_SHIFT)); 4566 BUILD_BUG_ON(SKB_GSO_UDP_L4 != (NETIF_F_GSO_UDP_L4 >> NETIF_F_GSO_SHIFT)); 4567 4568 return (features & feature) == feature; 4569 } 4570 4571 static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features) 4572 { 4573 return net_gso_ok(features, skb_shinfo(skb)->gso_type) && 4574 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST)); 4575 } 4576 4577 static inline bool netif_needs_gso(struct sk_buff *skb, 4578 netdev_features_t features) 4579 { 4580 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) || 4581 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) && 4582 (skb->ip_summed != CHECKSUM_UNNECESSARY))); 4583 } 4584 4585 static inline void netif_set_gso_max_size(struct net_device *dev, 4586 unsigned int size) 4587 { 4588 dev->gso_max_size = size; 4589 } 4590 4591 static inline void skb_gso_error_unwind(struct sk_buff *skb, __be16 protocol, 4592 int pulled_hlen, u16 mac_offset, 4593 int mac_len) 4594 { 4595 skb->protocol = protocol; 4596 skb->encapsulation = 1; 4597 skb_push(skb, pulled_hlen); 4598 skb_reset_transport_header(skb); 4599 skb->mac_header = mac_offset; 4600 skb->network_header = skb->mac_header + mac_len; 4601 skb->mac_len = mac_len; 4602 } 4603 4604 static inline bool netif_is_macsec(const struct net_device *dev) 4605 { 4606 return dev->priv_flags & IFF_MACSEC; 4607 } 4608 4609 static inline bool netif_is_macvlan(const struct net_device *dev) 4610 { 4611 return dev->priv_flags & IFF_MACVLAN; 4612 } 4613 4614 static inline bool netif_is_macvlan_port(const struct net_device *dev) 4615 { 4616 return dev->priv_flags & IFF_MACVLAN_PORT; 4617 } 4618 4619 static inline bool netif_is_bond_master(const struct net_device *dev) 4620 { 4621 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING; 4622 } 4623 4624 static inline bool netif_is_bond_slave(const struct net_device *dev) 4625 { 4626 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING; 4627 } 4628 4629 static inline bool netif_supports_nofcs(struct net_device *dev) 4630 { 4631 return dev->priv_flags & IFF_SUPP_NOFCS; 4632 } 4633 4634 static inline bool netif_has_l3_rx_handler(const struct net_device *dev) 4635 { 4636 return dev->priv_flags & IFF_L3MDEV_RX_HANDLER; 4637 } 4638 4639 static inline bool netif_is_l3_master(const struct net_device *dev) 4640 { 4641 return dev->priv_flags & IFF_L3MDEV_MASTER; 4642 } 4643 4644 static inline bool netif_is_l3_slave(const struct net_device *dev) 4645 { 4646 return dev->priv_flags & IFF_L3MDEV_SLAVE; 4647 } 4648 4649 static inline bool netif_is_bridge_master(const struct net_device *dev) 4650 { 4651 return dev->priv_flags & IFF_EBRIDGE; 4652 } 4653 4654 static inline bool netif_is_bridge_port(const struct net_device *dev) 4655 { 4656 return dev->priv_flags & IFF_BRIDGE_PORT; 4657 } 4658 4659 static inline bool netif_is_ovs_master(const struct net_device *dev) 4660 { 4661 return dev->priv_flags & IFF_OPENVSWITCH; 4662 } 4663 4664 static inline bool netif_is_ovs_port(const struct net_device *dev) 4665 { 4666 return dev->priv_flags & IFF_OVS_DATAPATH; 4667 } 4668 4669 static inline bool netif_is_team_master(const struct net_device *dev) 4670 { 4671 return dev->priv_flags & IFF_TEAM; 4672 } 4673 4674 static inline bool netif_is_team_port(const struct net_device *dev) 4675 { 4676 return dev->priv_flags & IFF_TEAM_PORT; 4677 } 4678 4679 static inline bool netif_is_lag_master(const struct net_device *dev) 4680 { 4681 return netif_is_bond_master(dev) || netif_is_team_master(dev); 4682 } 4683 4684 static inline bool netif_is_lag_port(const struct net_device *dev) 4685 { 4686 return netif_is_bond_slave(dev) || netif_is_team_port(dev); 4687 } 4688 4689 static inline bool netif_is_rxfh_configured(const struct net_device *dev) 4690 { 4691 return dev->priv_flags & IFF_RXFH_CONFIGURED; 4692 } 4693 4694 static inline bool netif_is_failover(const struct net_device *dev) 4695 { 4696 return dev->priv_flags & IFF_FAILOVER; 4697 } 4698 4699 static inline bool netif_is_failover_slave(const struct net_device *dev) 4700 { 4701 return dev->priv_flags & IFF_FAILOVER_SLAVE; 4702 } 4703 4704 /* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */ 4705 static inline void netif_keep_dst(struct net_device *dev) 4706 { 4707 dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM); 4708 } 4709 4710 /* return true if dev can't cope with mtu frames that need vlan tag insertion */ 4711 static inline bool netif_reduces_vlan_mtu(struct net_device *dev) 4712 { 4713 /* TODO: reserve and use an additional IFF bit, if we get more users */ 4714 return dev->priv_flags & IFF_MACSEC; 4715 } 4716 4717 extern struct pernet_operations __net_initdata loopback_net_ops; 4718 4719 /* Logging, debugging and troubleshooting/diagnostic helpers. */ 4720 4721 /* netdev_printk helpers, similar to dev_printk */ 4722 4723 static inline const char *netdev_name(const struct net_device *dev) 4724 { 4725 if (!dev->name[0] || strchr(dev->name, '%')) 4726 return "(unnamed net_device)"; 4727 return dev->name; 4728 } 4729 4730 static inline bool netdev_unregistering(const struct net_device *dev) 4731 { 4732 return dev->reg_state == NETREG_UNREGISTERING; 4733 } 4734 4735 static inline const char *netdev_reg_state(const struct net_device *dev) 4736 { 4737 switch (dev->reg_state) { 4738 case NETREG_UNINITIALIZED: return " (uninitialized)"; 4739 case NETREG_REGISTERED: return ""; 4740 case NETREG_UNREGISTERING: return " (unregistering)"; 4741 case NETREG_UNREGISTERED: return " (unregistered)"; 4742 case NETREG_RELEASED: return " (released)"; 4743 case NETREG_DUMMY: return " (dummy)"; 4744 } 4745 4746 WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, dev->reg_state); 4747 return " (unknown)"; 4748 } 4749 4750 __printf(3, 4) __cold 4751 void netdev_printk(const char *level, const struct net_device *dev, 4752 const char *format, ...); 4753 __printf(2, 3) __cold 4754 void netdev_emerg(const struct net_device *dev, const char *format, ...); 4755 __printf(2, 3) __cold 4756 void netdev_alert(const struct net_device *dev, const char *format, ...); 4757 __printf(2, 3) __cold 4758 void netdev_crit(const struct net_device *dev, const char *format, ...); 4759 __printf(2, 3) __cold 4760 void netdev_err(const struct net_device *dev, const char *format, ...); 4761 __printf(2, 3) __cold 4762 void netdev_warn(const struct net_device *dev, const char *format, ...); 4763 __printf(2, 3) __cold 4764 void netdev_notice(const struct net_device *dev, const char *format, ...); 4765 __printf(2, 3) __cold 4766 void netdev_info(const struct net_device *dev, const char *format, ...); 4767 4768 #define netdev_level_once(level, dev, fmt, ...) \ 4769 do { \ 4770 static bool __print_once __read_mostly; \ 4771 \ 4772 if (!__print_once) { \ 4773 __print_once = true; \ 4774 netdev_printk(level, dev, fmt, ##__VA_ARGS__); \ 4775 } \ 4776 } while (0) 4777 4778 #define netdev_emerg_once(dev, fmt, ...) \ 4779 netdev_level_once(KERN_EMERG, dev, fmt, ##__VA_ARGS__) 4780 #define netdev_alert_once(dev, fmt, ...) \ 4781 netdev_level_once(KERN_ALERT, dev, fmt, ##__VA_ARGS__) 4782 #define netdev_crit_once(dev, fmt, ...) \ 4783 netdev_level_once(KERN_CRIT, dev, fmt, ##__VA_ARGS__) 4784 #define netdev_err_once(dev, fmt, ...) \ 4785 netdev_level_once(KERN_ERR, dev, fmt, ##__VA_ARGS__) 4786 #define netdev_warn_once(dev, fmt, ...) \ 4787 netdev_level_once(KERN_WARNING, dev, fmt, ##__VA_ARGS__) 4788 #define netdev_notice_once(dev, fmt, ...) \ 4789 netdev_level_once(KERN_NOTICE, dev, fmt, ##__VA_ARGS__) 4790 #define netdev_info_once(dev, fmt, ...) \ 4791 netdev_level_once(KERN_INFO, dev, fmt, ##__VA_ARGS__) 4792 4793 #define MODULE_ALIAS_NETDEV(device) \ 4794 MODULE_ALIAS("netdev-" device) 4795 4796 #if defined(CONFIG_DYNAMIC_DEBUG) 4797 #define netdev_dbg(__dev, format, args...) \ 4798 do { \ 4799 dynamic_netdev_dbg(__dev, format, ##args); \ 4800 } while (0) 4801 #elif defined(DEBUG) 4802 #define netdev_dbg(__dev, format, args...) \ 4803 netdev_printk(KERN_DEBUG, __dev, format, ##args) 4804 #else 4805 #define netdev_dbg(__dev, format, args...) \ 4806 ({ \ 4807 if (0) \ 4808 netdev_printk(KERN_DEBUG, __dev, format, ##args); \ 4809 }) 4810 #endif 4811 4812 #if defined(VERBOSE_DEBUG) 4813 #define netdev_vdbg netdev_dbg 4814 #else 4815 4816 #define netdev_vdbg(dev, format, args...) \ 4817 ({ \ 4818 if (0) \ 4819 netdev_printk(KERN_DEBUG, dev, format, ##args); \ 4820 0; \ 4821 }) 4822 #endif 4823 4824 /* 4825 * netdev_WARN() acts like dev_printk(), but with the key difference 4826 * of using a WARN/WARN_ON to get the message out, including the 4827 * file/line information and a backtrace. 4828 */ 4829 #define netdev_WARN(dev, format, args...) \ 4830 WARN(1, "netdevice: %s%s: " format, netdev_name(dev), \ 4831 netdev_reg_state(dev), ##args) 4832 4833 #define netdev_WARN_ONCE(dev, format, args...) \ 4834 WARN_ONCE(1, "netdevice: %s%s: " format, netdev_name(dev), \ 4835 netdev_reg_state(dev), ##args) 4836 4837 /* netif printk helpers, similar to netdev_printk */ 4838 4839 #define netif_printk(priv, type, level, dev, fmt, args...) \ 4840 do { \ 4841 if (netif_msg_##type(priv)) \ 4842 netdev_printk(level, (dev), fmt, ##args); \ 4843 } while (0) 4844 4845 #define netif_level(level, priv, type, dev, fmt, args...) \ 4846 do { \ 4847 if (netif_msg_##type(priv)) \ 4848 netdev_##level(dev, fmt, ##args); \ 4849 } while (0) 4850 4851 #define netif_emerg(priv, type, dev, fmt, args...) \ 4852 netif_level(emerg, priv, type, dev, fmt, ##args) 4853 #define netif_alert(priv, type, dev, fmt, args...) \ 4854 netif_level(alert, priv, type, dev, fmt, ##args) 4855 #define netif_crit(priv, type, dev, fmt, args...) \ 4856 netif_level(crit, priv, type, dev, fmt, ##args) 4857 #define netif_err(priv, type, dev, fmt, args...) \ 4858 netif_level(err, priv, type, dev, fmt, ##args) 4859 #define netif_warn(priv, type, dev, fmt, args...) \ 4860 netif_level(warn, priv, type, dev, fmt, ##args) 4861 #define netif_notice(priv, type, dev, fmt, args...) \ 4862 netif_level(notice, priv, type, dev, fmt, ##args) 4863 #define netif_info(priv, type, dev, fmt, args...) \ 4864 netif_level(info, priv, type, dev, fmt, ##args) 4865 4866 #if defined(CONFIG_DYNAMIC_DEBUG) 4867 #define netif_dbg(priv, type, netdev, format, args...) \ 4868 do { \ 4869 if (netif_msg_##type(priv)) \ 4870 dynamic_netdev_dbg(netdev, format, ##args); \ 4871 } while (0) 4872 #elif defined(DEBUG) 4873 #define netif_dbg(priv, type, dev, format, args...) \ 4874 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args) 4875 #else 4876 #define netif_dbg(priv, type, dev, format, args...) \ 4877 ({ \ 4878 if (0) \ 4879 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \ 4880 0; \ 4881 }) 4882 #endif 4883 4884 /* if @cond then downgrade to debug, else print at @level */ 4885 #define netif_cond_dbg(priv, type, netdev, cond, level, fmt, args...) \ 4886 do { \ 4887 if (cond) \ 4888 netif_dbg(priv, type, netdev, fmt, ##args); \ 4889 else \ 4890 netif_ ## level(priv, type, netdev, fmt, ##args); \ 4891 } while (0) 4892 4893 #if defined(VERBOSE_DEBUG) 4894 #define netif_vdbg netif_dbg 4895 #else 4896 #define netif_vdbg(priv, type, dev, format, args...) \ 4897 ({ \ 4898 if (0) \ 4899 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \ 4900 0; \ 4901 }) 4902 #endif 4903 4904 /* 4905 * The list of packet types we will receive (as opposed to discard) 4906 * and the routines to invoke. 4907 * 4908 * Why 16. Because with 16 the only overlap we get on a hash of the 4909 * low nibble of the protocol value is RARP/SNAP/X.25. 4910 * 4911 * 0800 IP 4912 * 0001 802.3 4913 * 0002 AX.25 4914 * 0004 802.2 4915 * 8035 RARP 4916 * 0005 SNAP 4917 * 0805 X.25 4918 * 0806 ARP 4919 * 8137 IPX 4920 * 0009 Localtalk 4921 * 86DD IPv6 4922 */ 4923 #define PTYPE_HASH_SIZE (16) 4924 #define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1) 4925 4926 extern struct net_device *blackhole_netdev; 4927 4928 #endif /* _LINUX_NETDEVICE_H */ 4929