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