1 /* 2 * INET An implementation of the TCP/IP protocol suite for the LINUX 3 * operating system. INET is implemented using the BSD Socket 4 * interface as the means of communication with the user level. 5 * 6 * Definitions for the Interfaces handler. 7 * 8 * Version: @(#)dev.h 1.0.10 08/12/93 9 * 10 * Authors: Ross Biro 11 * Fred N. van Kempen, <[email protected]> 12 * Corey Minyard <[email protected]> 13 * Donald J. Becker, <[email protected]> 14 * Alan Cox, <[email protected]> 15 * Bjorn Ekwall. <[email protected]> 16 * Pekka Riikonen <[email protected]> 17 * 18 * This program is free software; you can redistribute it and/or 19 * modify it under the terms of the GNU General Public License 20 * as published by the Free Software Foundation; either version 21 * 2 of the License, or (at your option) any later version. 22 * 23 * Moved to /usr/include/linux for NET3 24 */ 25 #ifndef _LINUX_NETDEVICE_H 26 #define _LINUX_NETDEVICE_H 27 28 #include <linux/timer.h> 29 #include <linux/bug.h> 30 #include <linux/delay.h> 31 #include <linux/atomic.h> 32 #include <linux/prefetch.h> 33 #include <asm/cache.h> 34 #include <asm/byteorder.h> 35 36 #include <linux/percpu.h> 37 #include <linux/rculist.h> 38 #include <linux/workqueue.h> 39 #include <linux/dynamic_queue_limits.h> 40 41 #include <linux/ethtool.h> 42 #include <net/net_namespace.h> 43 #ifdef CONFIG_DCB 44 #include <net/dcbnl.h> 45 #endif 46 #include <net/netprio_cgroup.h> 47 #include <net/xdp.h> 48 49 #include <linux/netdev_features.h> 50 #include <linux/neighbour.h> 51 #include <uapi/linux/netdevice.h> 52 #include <uapi/linux/if_bonding.h> 53 #include <uapi/linux/pkt_cls.h> 54 #include <linux/hashtable.h> 55 56 struct netpoll_info; 57 struct device; 58 struct phy_device; 59 struct dsa_port; 60 61 struct sfp_bus; 62 /* 802.11 specific */ 63 struct wireless_dev; 64 /* 802.15.4 specific */ 65 struct wpan_dev; 66 struct mpls_dev; 67 /* UDP Tunnel offloads */ 68 struct udp_tunnel_info; 69 struct bpf_prog; 70 struct xdp_buff; 71 72 void netdev_set_default_ethtool_ops(struct net_device *dev, 73 const struct ethtool_ops *ops); 74 75 /* Backlog congestion levels */ 76 #define NET_RX_SUCCESS 0 /* keep 'em coming, baby */ 77 #define NET_RX_DROP 1 /* packet dropped */ 78 79 /* 80 * Transmit return codes: transmit return codes originate from three different 81 * namespaces: 82 * 83 * - qdisc return codes 84 * - driver transmit return codes 85 * - errno values 86 * 87 * Drivers are allowed to return any one of those in their hard_start_xmit() 88 * function. Real network devices commonly used with qdiscs should only return 89 * the driver transmit return codes though - when qdiscs are used, the actual 90 * transmission happens asynchronously, so the value is not propagated to 91 * higher layers. Virtual network devices transmit synchronously; in this case 92 * the driver transmit return codes are consumed by dev_queue_xmit(), and all 93 * others are propagated to higher layers. 94 */ 95 96 /* qdisc ->enqueue() return codes. */ 97 #define NET_XMIT_SUCCESS 0x00 98 #define NET_XMIT_DROP 0x01 /* skb dropped */ 99 #define NET_XMIT_CN 0x02 /* congestion notification */ 100 #define NET_XMIT_MASK 0x0f /* qdisc flags in net/sch_generic.h */ 101 102 /* NET_XMIT_CN is special. It does not guarantee that this packet is lost. It 103 * indicates that the device will soon be dropping packets, or already drops 104 * some packets of the same priority; prompting us to send less aggressively. */ 105 #define net_xmit_eval(e) ((e) == NET_XMIT_CN ? 0 : (e)) 106 #define net_xmit_errno(e) ((e) != NET_XMIT_CN ? -ENOBUFS : 0) 107 108 /* Driver transmit return codes */ 109 #define NETDEV_TX_MASK 0xf0 110 111 enum netdev_tx { 112 __NETDEV_TX_MIN = INT_MIN, /* make sure enum is signed */ 113 NETDEV_TX_OK = 0x00, /* driver took care of packet */ 114 NETDEV_TX_BUSY = 0x10, /* driver tx path was busy*/ 115 }; 116 typedef enum netdev_tx netdev_tx_t; 117 118 /* 119 * Current order: NETDEV_TX_MASK > NET_XMIT_MASK >= 0 is significant; 120 * hard_start_xmit() return < NET_XMIT_MASK means skb was consumed. 121 */ 122 static inline bool dev_xmit_complete(int rc) 123 { 124 /* 125 * Positive cases with an skb consumed by a driver: 126 * - successful transmission (rc == NETDEV_TX_OK) 127 * - error while transmitting (rc < 0) 128 * - error while queueing to a different device (rc & NET_XMIT_MASK) 129 */ 130 if (likely(rc < NET_XMIT_MASK)) 131 return true; 132 133 return false; 134 } 135 136 /* 137 * Compute the worst-case header length according to the protocols 138 * used. 139 */ 140 141 #if defined(CONFIG_HYPERV_NET) 142 # define LL_MAX_HEADER 128 143 #elif defined(CONFIG_WLAN) || IS_ENABLED(CONFIG_AX25) 144 # if defined(CONFIG_MAC80211_MESH) 145 # define LL_MAX_HEADER 128 146 # else 147 # define LL_MAX_HEADER 96 148 # endif 149 #else 150 # define LL_MAX_HEADER 32 151 #endif 152 153 #if !IS_ENABLED(CONFIG_NET_IPIP) && !IS_ENABLED(CONFIG_NET_IPGRE) && \ 154 !IS_ENABLED(CONFIG_IPV6_SIT) && !IS_ENABLED(CONFIG_IPV6_TUNNEL) 155 #define MAX_HEADER LL_MAX_HEADER 156 #else 157 #define MAX_HEADER (LL_MAX_HEADER + 48) 158 #endif 159 160 /* 161 * Old network device statistics. Fields are native words 162 * (unsigned long) so they can be read and written atomically. 163 */ 164 165 struct net_device_stats { 166 unsigned long rx_packets; 167 unsigned long tx_packets; 168 unsigned long rx_bytes; 169 unsigned long tx_bytes; 170 unsigned long rx_errors; 171 unsigned long tx_errors; 172 unsigned long rx_dropped; 173 unsigned long tx_dropped; 174 unsigned long multicast; 175 unsigned long collisions; 176 unsigned long rx_length_errors; 177 unsigned long rx_over_errors; 178 unsigned long rx_crc_errors; 179 unsigned long rx_frame_errors; 180 unsigned long rx_fifo_errors; 181 unsigned long rx_missed_errors; 182 unsigned long tx_aborted_errors; 183 unsigned long tx_carrier_errors; 184 unsigned long tx_fifo_errors; 185 unsigned long tx_heartbeat_errors; 186 unsigned long tx_window_errors; 187 unsigned long rx_compressed; 188 unsigned long tx_compressed; 189 }; 190 191 192 #include <linux/cache.h> 193 #include <linux/skbuff.h> 194 195 #ifdef CONFIG_RPS 196 #include <linux/static_key.h> 197 extern struct static_key rps_needed; 198 extern struct static_key rfs_needed; 199 #endif 200 201 struct neighbour; 202 struct neigh_parms; 203 struct sk_buff; 204 205 struct netdev_hw_addr { 206 struct list_head list; 207 unsigned char addr[MAX_ADDR_LEN]; 208 unsigned char type; 209 #define NETDEV_HW_ADDR_T_LAN 1 210 #define NETDEV_HW_ADDR_T_SAN 2 211 #define NETDEV_HW_ADDR_T_SLAVE 3 212 #define NETDEV_HW_ADDR_T_UNICAST 4 213 #define NETDEV_HW_ADDR_T_MULTICAST 5 214 bool global_use; 215 int sync_cnt; 216 int refcount; 217 int synced; 218 struct rcu_head rcu_head; 219 }; 220 221 struct netdev_hw_addr_list { 222 struct list_head list; 223 int count; 224 }; 225 226 #define netdev_hw_addr_list_count(l) ((l)->count) 227 #define netdev_hw_addr_list_empty(l) (netdev_hw_addr_list_count(l) == 0) 228 #define netdev_hw_addr_list_for_each(ha, l) \ 229 list_for_each_entry(ha, &(l)->list, list) 230 231 #define netdev_uc_count(dev) netdev_hw_addr_list_count(&(dev)->uc) 232 #define netdev_uc_empty(dev) netdev_hw_addr_list_empty(&(dev)->uc) 233 #define netdev_for_each_uc_addr(ha, dev) \ 234 netdev_hw_addr_list_for_each(ha, &(dev)->uc) 235 236 #define netdev_mc_count(dev) netdev_hw_addr_list_count(&(dev)->mc) 237 #define netdev_mc_empty(dev) netdev_hw_addr_list_empty(&(dev)->mc) 238 #define netdev_for_each_mc_addr(ha, dev) \ 239 netdev_hw_addr_list_for_each(ha, &(dev)->mc) 240 241 struct hh_cache { 242 unsigned int hh_len; 243 seqlock_t hh_lock; 244 245 /* cached hardware header; allow for machine alignment needs. */ 246 #define HH_DATA_MOD 16 247 #define HH_DATA_OFF(__len) \ 248 (HH_DATA_MOD - (((__len - 1) & (HH_DATA_MOD - 1)) + 1)) 249 #define HH_DATA_ALIGN(__len) \ 250 (((__len)+(HH_DATA_MOD-1))&~(HH_DATA_MOD - 1)) 251 unsigned long hh_data[HH_DATA_ALIGN(LL_MAX_HEADER) / sizeof(long)]; 252 }; 253 254 /* Reserve HH_DATA_MOD byte-aligned hard_header_len, but at least that much. 255 * Alternative is: 256 * dev->hard_header_len ? (dev->hard_header_len + 257 * (HH_DATA_MOD - 1)) & ~(HH_DATA_MOD - 1) : 0 258 * 259 * We could use other alignment values, but we must maintain the 260 * relationship HH alignment <= LL alignment. 261 */ 262 #define LL_RESERVED_SPACE(dev) \ 263 ((((dev)->hard_header_len+(dev)->needed_headroom)&~(HH_DATA_MOD - 1)) + HH_DATA_MOD) 264 #define LL_RESERVED_SPACE_EXTRA(dev,extra) \ 265 ((((dev)->hard_header_len+(dev)->needed_headroom+(extra))&~(HH_DATA_MOD - 1)) + HH_DATA_MOD) 266 267 struct header_ops { 268 int (*create) (struct sk_buff *skb, struct net_device *dev, 269 unsigned short type, const void *daddr, 270 const void *saddr, unsigned int len); 271 int (*parse)(const struct sk_buff *skb, unsigned char *haddr); 272 int (*cache)(const struct neighbour *neigh, struct hh_cache *hh, __be16 type); 273 void (*cache_update)(struct hh_cache *hh, 274 const struct net_device *dev, 275 const unsigned char *haddr); 276 bool (*validate)(const char *ll_header, unsigned int len); 277 }; 278 279 /* These flag bits are private to the generic network queueing 280 * layer; they may not be explicitly referenced by any other 281 * code. 282 */ 283 284 enum netdev_state_t { 285 __LINK_STATE_START, 286 __LINK_STATE_PRESENT, 287 __LINK_STATE_NOCARRIER, 288 __LINK_STATE_LINKWATCH_PENDING, 289 __LINK_STATE_DORMANT, 290 }; 291 292 293 /* 294 * This structure holds boot-time configured netdevice settings. They 295 * are then used in the device probing. 296 */ 297 struct netdev_boot_setup { 298 char name[IFNAMSIZ]; 299 struct ifmap map; 300 }; 301 #define NETDEV_BOOT_SETUP_MAX 8 302 303 int __init netdev_boot_setup(char *str); 304 305 struct gro_list { 306 struct list_head list; 307 int count; 308 }; 309 310 /* 311 * size of gro hash buckets, must less than bit number of 312 * napi_struct::gro_bitmask 313 */ 314 #define GRO_HASH_BUCKETS 8 315 316 /* 317 * Structure for NAPI scheduling similar to tasklet but with weighting 318 */ 319 struct napi_struct { 320 /* The poll_list must only be managed by the entity which 321 * changes the state of the NAPI_STATE_SCHED bit. This means 322 * whoever atomically sets that bit can add this napi_struct 323 * to the per-CPU poll_list, and whoever clears that bit 324 * can remove from the list right before clearing the bit. 325 */ 326 struct list_head poll_list; 327 328 unsigned long state; 329 int weight; 330 unsigned long gro_bitmask; 331 int (*poll)(struct napi_struct *, int); 332 #ifdef CONFIG_NETPOLL 333 int poll_owner; 334 #endif 335 struct net_device *dev; 336 struct gro_list gro_hash[GRO_HASH_BUCKETS]; 337 struct sk_buff *skb; 338 struct hrtimer timer; 339 struct list_head dev_list; 340 struct hlist_node napi_hash_node; 341 unsigned int napi_id; 342 }; 343 344 enum { 345 NAPI_STATE_SCHED, /* Poll is scheduled */ 346 NAPI_STATE_MISSED, /* reschedule a napi */ 347 NAPI_STATE_DISABLE, /* Disable pending */ 348 NAPI_STATE_NPSVC, /* Netpoll - don't dequeue from poll_list */ 349 NAPI_STATE_HASHED, /* In NAPI hash (busy polling possible) */ 350 NAPI_STATE_NO_BUSY_POLL,/* Do not add in napi_hash, no busy polling */ 351 NAPI_STATE_IN_BUSY_POLL,/* sk_busy_loop() owns this NAPI */ 352 }; 353 354 enum { 355 NAPIF_STATE_SCHED = BIT(NAPI_STATE_SCHED), 356 NAPIF_STATE_MISSED = BIT(NAPI_STATE_MISSED), 357 NAPIF_STATE_DISABLE = BIT(NAPI_STATE_DISABLE), 358 NAPIF_STATE_NPSVC = BIT(NAPI_STATE_NPSVC), 359 NAPIF_STATE_HASHED = BIT(NAPI_STATE_HASHED), 360 NAPIF_STATE_NO_BUSY_POLL = BIT(NAPI_STATE_NO_BUSY_POLL), 361 NAPIF_STATE_IN_BUSY_POLL = BIT(NAPI_STATE_IN_BUSY_POLL), 362 }; 363 364 enum gro_result { 365 GRO_MERGED, 366 GRO_MERGED_FREE, 367 GRO_HELD, 368 GRO_NORMAL, 369 GRO_DROP, 370 GRO_CONSUMED, 371 }; 372 typedef enum gro_result gro_result_t; 373 374 /* 375 * enum rx_handler_result - Possible return values for rx_handlers. 376 * @RX_HANDLER_CONSUMED: skb was consumed by rx_handler, do not process it 377 * further. 378 * @RX_HANDLER_ANOTHER: Do another round in receive path. This is indicated in 379 * case skb->dev was changed by rx_handler. 380 * @RX_HANDLER_EXACT: Force exact delivery, no wildcard. 381 * @RX_HANDLER_PASS: Do nothing, pass the skb as if no rx_handler was called. 382 * 383 * rx_handlers are functions called from inside __netif_receive_skb(), to do 384 * special processing of the skb, prior to delivery to protocol handlers. 385 * 386 * Currently, a net_device can only have a single rx_handler registered. Trying 387 * to register a second rx_handler will return -EBUSY. 388 * 389 * To register a rx_handler on a net_device, use netdev_rx_handler_register(). 390 * To unregister a rx_handler on a net_device, use 391 * netdev_rx_handler_unregister(). 392 * 393 * Upon return, rx_handler is expected to tell __netif_receive_skb() what to 394 * do with the skb. 395 * 396 * If the rx_handler consumed the skb in some way, it should return 397 * RX_HANDLER_CONSUMED. This is appropriate when the rx_handler arranged for 398 * the skb to be delivered in some other way. 399 * 400 * If the rx_handler changed skb->dev, to divert the skb to another 401 * net_device, it should return RX_HANDLER_ANOTHER. The rx_handler for the 402 * new device will be called if it exists. 403 * 404 * If the rx_handler decides the skb should be ignored, it should return 405 * RX_HANDLER_EXACT. The skb will only be delivered to protocol handlers that 406 * are registered on exact device (ptype->dev == skb->dev). 407 * 408 * If the rx_handler didn't change skb->dev, but wants the skb to be normally 409 * delivered, it should return RX_HANDLER_PASS. 410 * 411 * A device without a registered rx_handler will behave as if rx_handler 412 * returned RX_HANDLER_PASS. 413 */ 414 415 enum rx_handler_result { 416 RX_HANDLER_CONSUMED, 417 RX_HANDLER_ANOTHER, 418 RX_HANDLER_EXACT, 419 RX_HANDLER_PASS, 420 }; 421 typedef enum rx_handler_result rx_handler_result_t; 422 typedef rx_handler_result_t rx_handler_func_t(struct sk_buff **pskb); 423 424 void __napi_schedule(struct napi_struct *n); 425 void __napi_schedule_irqoff(struct napi_struct *n); 426 427 static inline bool napi_disable_pending(struct napi_struct *n) 428 { 429 return test_bit(NAPI_STATE_DISABLE, &n->state); 430 } 431 432 bool napi_schedule_prep(struct napi_struct *n); 433 434 /** 435 * napi_schedule - schedule NAPI poll 436 * @n: NAPI context 437 * 438 * Schedule NAPI poll routine to be called if it is not already 439 * running. 440 */ 441 static inline void napi_schedule(struct napi_struct *n) 442 { 443 if (napi_schedule_prep(n)) 444 __napi_schedule(n); 445 } 446 447 /** 448 * napi_schedule_irqoff - schedule NAPI poll 449 * @n: NAPI context 450 * 451 * Variant of napi_schedule(), assuming hard irqs are masked. 452 */ 453 static inline void napi_schedule_irqoff(struct napi_struct *n) 454 { 455 if (napi_schedule_prep(n)) 456 __napi_schedule_irqoff(n); 457 } 458 459 /* Try to reschedule poll. Called by dev->poll() after napi_complete(). */ 460 static inline bool napi_reschedule(struct napi_struct *napi) 461 { 462 if (napi_schedule_prep(napi)) { 463 __napi_schedule(napi); 464 return true; 465 } 466 return false; 467 } 468 469 bool napi_complete_done(struct napi_struct *n, int work_done); 470 /** 471 * napi_complete - NAPI processing complete 472 * @n: NAPI context 473 * 474 * Mark NAPI processing as complete. 475 * Consider using napi_complete_done() instead. 476 * Return false if device should avoid rearming interrupts. 477 */ 478 static inline bool napi_complete(struct napi_struct *n) 479 { 480 return napi_complete_done(n, 0); 481 } 482 483 /** 484 * napi_hash_del - remove a NAPI from global table 485 * @napi: NAPI context 486 * 487 * Warning: caller must observe RCU grace period 488 * before freeing memory containing @napi, if 489 * this function returns true. 490 * Note: core networking stack automatically calls it 491 * from netif_napi_del(). 492 * Drivers might want to call this helper to combine all 493 * the needed RCU grace periods into a single one. 494 */ 495 bool napi_hash_del(struct napi_struct *napi); 496 497 /** 498 * napi_disable - prevent NAPI from scheduling 499 * @n: NAPI context 500 * 501 * Stop NAPI from being scheduled on this context. 502 * Waits till any outstanding processing completes. 503 */ 504 void napi_disable(struct napi_struct *n); 505 506 /** 507 * napi_enable - enable NAPI scheduling 508 * @n: NAPI context 509 * 510 * Resume NAPI from being scheduled on this context. 511 * Must be paired with napi_disable. 512 */ 513 static inline void napi_enable(struct napi_struct *n) 514 { 515 BUG_ON(!test_bit(NAPI_STATE_SCHED, &n->state)); 516 smp_mb__before_atomic(); 517 clear_bit(NAPI_STATE_SCHED, &n->state); 518 clear_bit(NAPI_STATE_NPSVC, &n->state); 519 } 520 521 /** 522 * napi_synchronize - wait until NAPI is not running 523 * @n: NAPI context 524 * 525 * Wait until NAPI is done being scheduled on this context. 526 * Waits till any outstanding processing completes but 527 * does not disable future activations. 528 */ 529 static inline void napi_synchronize(const struct napi_struct *n) 530 { 531 if (IS_ENABLED(CONFIG_SMP)) 532 while (test_bit(NAPI_STATE_SCHED, &n->state)) 533 msleep(1); 534 else 535 barrier(); 536 } 537 538 /** 539 * napi_if_scheduled_mark_missed - if napi is running, set the 540 * NAPIF_STATE_MISSED 541 * @n: NAPI context 542 * 543 * If napi is running, set the NAPIF_STATE_MISSED, and return true if 544 * NAPI is scheduled. 545 **/ 546 static inline bool napi_if_scheduled_mark_missed(struct napi_struct *n) 547 { 548 unsigned long val, new; 549 550 do { 551 val = READ_ONCE(n->state); 552 if (val & NAPIF_STATE_DISABLE) 553 return true; 554 555 if (!(val & NAPIF_STATE_SCHED)) 556 return false; 557 558 new = val | NAPIF_STATE_MISSED; 559 } while (cmpxchg(&n->state, val, new) != val); 560 561 return true; 562 } 563 564 enum netdev_queue_state_t { 565 __QUEUE_STATE_DRV_XOFF, 566 __QUEUE_STATE_STACK_XOFF, 567 __QUEUE_STATE_FROZEN, 568 }; 569 570 #define QUEUE_STATE_DRV_XOFF (1 << __QUEUE_STATE_DRV_XOFF) 571 #define QUEUE_STATE_STACK_XOFF (1 << __QUEUE_STATE_STACK_XOFF) 572 #define QUEUE_STATE_FROZEN (1 << __QUEUE_STATE_FROZEN) 573 574 #define QUEUE_STATE_ANY_XOFF (QUEUE_STATE_DRV_XOFF | QUEUE_STATE_STACK_XOFF) 575 #define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \ 576 QUEUE_STATE_FROZEN) 577 #define QUEUE_STATE_DRV_XOFF_OR_FROZEN (QUEUE_STATE_DRV_XOFF | \ 578 QUEUE_STATE_FROZEN) 579 580 /* 581 * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue. The 582 * netif_tx_* functions below are used to manipulate this flag. The 583 * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit 584 * queue independently. The netif_xmit_*stopped functions below are called 585 * to check if the queue has been stopped by the driver or stack (either 586 * of the XOFF bits are set in the state). Drivers should not need to call 587 * netif_xmit*stopped functions, they should only be using netif_tx_*. 588 */ 589 590 struct netdev_queue { 591 /* 592 * read-mostly part 593 */ 594 struct net_device *dev; 595 struct Qdisc __rcu *qdisc; 596 struct Qdisc *qdisc_sleeping; 597 #ifdef CONFIG_SYSFS 598 struct kobject kobj; 599 #endif 600 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA) 601 int numa_node; 602 #endif 603 unsigned long tx_maxrate; 604 /* 605 * Number of TX timeouts for this queue 606 * (/sys/class/net/DEV/Q/trans_timeout) 607 */ 608 unsigned long trans_timeout; 609 610 /* Subordinate device that the queue has been assigned to */ 611 struct net_device *sb_dev; 612 #ifdef CONFIG_XDP_SOCKETS 613 struct xdp_umem *umem; 614 #endif 615 /* 616 * write-mostly part 617 */ 618 spinlock_t _xmit_lock ____cacheline_aligned_in_smp; 619 int xmit_lock_owner; 620 /* 621 * Time (in jiffies) of last Tx 622 */ 623 unsigned long trans_start; 624 625 unsigned long state; 626 627 #ifdef CONFIG_BQL 628 struct dql dql; 629 #endif 630 } ____cacheline_aligned_in_smp; 631 632 extern int sysctl_fb_tunnels_only_for_init_net; 633 extern int sysctl_devconf_inherit_init_net; 634 635 static inline bool net_has_fallback_tunnels(const struct net *net) 636 { 637 return net == &init_net || 638 !IS_ENABLED(CONFIG_SYSCTL) || 639 !sysctl_fb_tunnels_only_for_init_net; 640 } 641 642 static inline int netdev_queue_numa_node_read(const struct netdev_queue *q) 643 { 644 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA) 645 return q->numa_node; 646 #else 647 return NUMA_NO_NODE; 648 #endif 649 } 650 651 static inline void netdev_queue_numa_node_write(struct netdev_queue *q, int node) 652 { 653 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA) 654 q->numa_node = node; 655 #endif 656 } 657 658 #ifdef CONFIG_RPS 659 /* 660 * This structure holds an RPS map which can be of variable length. The 661 * map is an array of CPUs. 662 */ 663 struct rps_map { 664 unsigned int len; 665 struct rcu_head rcu; 666 u16 cpus[0]; 667 }; 668 #define RPS_MAP_SIZE(_num) (sizeof(struct rps_map) + ((_num) * sizeof(u16))) 669 670 /* 671 * The rps_dev_flow structure contains the mapping of a flow to a CPU, the 672 * tail pointer for that CPU's input queue at the time of last enqueue, and 673 * a hardware filter index. 674 */ 675 struct rps_dev_flow { 676 u16 cpu; 677 u16 filter; 678 unsigned int last_qtail; 679 }; 680 #define RPS_NO_FILTER 0xffff 681 682 /* 683 * The rps_dev_flow_table structure contains a table of flow mappings. 684 */ 685 struct rps_dev_flow_table { 686 unsigned int mask; 687 struct rcu_head rcu; 688 struct rps_dev_flow flows[0]; 689 }; 690 #define RPS_DEV_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_dev_flow_table) + \ 691 ((_num) * sizeof(struct rps_dev_flow))) 692 693 /* 694 * The rps_sock_flow_table contains mappings of flows to the last CPU 695 * on which they were processed by the application (set in recvmsg). 696 * Each entry is a 32bit value. Upper part is the high-order bits 697 * of flow hash, lower part is CPU number. 698 * rps_cpu_mask is used to partition the space, depending on number of 699 * possible CPUs : rps_cpu_mask = roundup_pow_of_two(nr_cpu_ids) - 1 700 * For example, if 64 CPUs are possible, rps_cpu_mask = 0x3f, 701 * meaning we use 32-6=26 bits for the hash. 702 */ 703 struct rps_sock_flow_table { 704 u32 mask; 705 706 u32 ents[0] ____cacheline_aligned_in_smp; 707 }; 708 #define RPS_SOCK_FLOW_TABLE_SIZE(_num) (offsetof(struct rps_sock_flow_table, ents[_num])) 709 710 #define RPS_NO_CPU 0xffff 711 712 extern u32 rps_cpu_mask; 713 extern struct rps_sock_flow_table __rcu *rps_sock_flow_table; 714 715 static inline void rps_record_sock_flow(struct rps_sock_flow_table *table, 716 u32 hash) 717 { 718 if (table && hash) { 719 unsigned int index = hash & table->mask; 720 u32 val = hash & ~rps_cpu_mask; 721 722 /* We only give a hint, preemption can change CPU under us */ 723 val |= raw_smp_processor_id(); 724 725 if (table->ents[index] != val) 726 table->ents[index] = val; 727 } 728 } 729 730 #ifdef CONFIG_RFS_ACCEL 731 bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index, u32 flow_id, 732 u16 filter_id); 733 #endif 734 #endif /* CONFIG_RPS */ 735 736 /* This structure contains an instance of an RX queue. */ 737 struct netdev_rx_queue { 738 #ifdef CONFIG_RPS 739 struct rps_map __rcu *rps_map; 740 struct rps_dev_flow_table __rcu *rps_flow_table; 741 #endif 742 struct kobject kobj; 743 struct net_device *dev; 744 struct xdp_rxq_info xdp_rxq; 745 #ifdef CONFIG_XDP_SOCKETS 746 struct xdp_umem *umem; 747 #endif 748 } ____cacheline_aligned_in_smp; 749 750 /* 751 * RX queue sysfs structures and functions. 752 */ 753 struct rx_queue_attribute { 754 struct attribute attr; 755 ssize_t (*show)(struct netdev_rx_queue *queue, char *buf); 756 ssize_t (*store)(struct netdev_rx_queue *queue, 757 const char *buf, size_t len); 758 }; 759 760 #ifdef CONFIG_XPS 761 /* 762 * This structure holds an XPS map which can be of variable length. The 763 * map is an array of queues. 764 */ 765 struct xps_map { 766 unsigned int len; 767 unsigned int alloc_len; 768 struct rcu_head rcu; 769 u16 queues[0]; 770 }; 771 #define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16))) 772 #define XPS_MIN_MAP_ALLOC ((L1_CACHE_ALIGN(offsetof(struct xps_map, queues[1])) \ 773 - sizeof(struct xps_map)) / sizeof(u16)) 774 775 /* 776 * This structure holds all XPS maps for device. Maps are indexed by CPU. 777 */ 778 struct xps_dev_maps { 779 struct rcu_head rcu; 780 struct xps_map __rcu *attr_map[0]; /* Either CPUs map or RXQs map */ 781 }; 782 783 #define XPS_CPU_DEV_MAPS_SIZE(_tcs) (sizeof(struct xps_dev_maps) + \ 784 (nr_cpu_ids * (_tcs) * sizeof(struct xps_map *))) 785 786 #define XPS_RXQ_DEV_MAPS_SIZE(_tcs, _rxqs) (sizeof(struct xps_dev_maps) +\ 787 (_rxqs * (_tcs) * sizeof(struct xps_map *))) 788 789 #endif /* CONFIG_XPS */ 790 791 #define TC_MAX_QUEUE 16 792 #define TC_BITMASK 15 793 /* HW offloaded queuing disciplines txq count and offset maps */ 794 struct netdev_tc_txq { 795 u16 count; 796 u16 offset; 797 }; 798 799 #if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE) 800 /* 801 * This structure is to hold information about the device 802 * configured to run FCoE protocol stack. 803 */ 804 struct netdev_fcoe_hbainfo { 805 char manufacturer[64]; 806 char serial_number[64]; 807 char hardware_version[64]; 808 char driver_version[64]; 809 char optionrom_version[64]; 810 char firmware_version[64]; 811 char model[256]; 812 char model_description[256]; 813 }; 814 #endif 815 816 #define MAX_PHYS_ITEM_ID_LEN 32 817 818 /* This structure holds a unique identifier to identify some 819 * physical item (port for example) used by a netdevice. 820 */ 821 struct netdev_phys_item_id { 822 unsigned char id[MAX_PHYS_ITEM_ID_LEN]; 823 unsigned char id_len; 824 }; 825 826 static inline bool netdev_phys_item_id_same(struct netdev_phys_item_id *a, 827 struct netdev_phys_item_id *b) 828 { 829 return a->id_len == b->id_len && 830 memcmp(a->id, b->id, a->id_len) == 0; 831 } 832 833 typedef u16 (*select_queue_fallback_t)(struct net_device *dev, 834 struct sk_buff *skb, 835 struct net_device *sb_dev); 836 837 enum tc_setup_type { 838 TC_SETUP_QDISC_MQPRIO, 839 TC_SETUP_CLSU32, 840 TC_SETUP_CLSFLOWER, 841 TC_SETUP_CLSMATCHALL, 842 TC_SETUP_CLSBPF, 843 TC_SETUP_BLOCK, 844 TC_SETUP_QDISC_CBS, 845 TC_SETUP_QDISC_RED, 846 TC_SETUP_QDISC_PRIO, 847 TC_SETUP_QDISC_MQ, 848 TC_SETUP_QDISC_ETF, 849 TC_SETUP_ROOT_QDISC, 850 TC_SETUP_QDISC_GRED, 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_QUERY_XSK_UMEM, 872 XDP_SETUP_XSK_UMEM, 873 }; 874 875 struct bpf_prog_offload_ops; 876 struct netlink_ext_ack; 877 struct xdp_umem; 878 879 struct netdev_bpf { 880 enum bpf_netdev_command command; 881 union { 882 /* XDP_SETUP_PROG */ 883 struct { 884 u32 flags; 885 struct bpf_prog *prog; 886 struct netlink_ext_ack *extack; 887 }; 888 /* XDP_QUERY_PROG, XDP_QUERY_PROG_HW */ 889 struct { 890 u32 prog_id; 891 /* flags with which program was installed */ 892 u32 prog_flags; 893 }; 894 /* BPF_OFFLOAD_MAP_ALLOC, BPF_OFFLOAD_MAP_FREE */ 895 struct { 896 struct bpf_offloaded_map *offmap; 897 }; 898 /* XDP_QUERY_XSK_UMEM, XDP_SETUP_XSK_UMEM */ 899 struct { 900 struct xdp_umem *umem; /* out for query*/ 901 u16 queue_id; /* in for query */ 902 } xsk; 903 }; 904 }; 905 906 #ifdef CONFIG_XFRM_OFFLOAD 907 struct xfrmdev_ops { 908 int (*xdo_dev_state_add) (struct xfrm_state *x); 909 void (*xdo_dev_state_delete) (struct xfrm_state *x); 910 void (*xdo_dev_state_free) (struct xfrm_state *x); 911 bool (*xdo_dev_offload_ok) (struct sk_buff *skb, 912 struct xfrm_state *x); 913 void (*xdo_dev_state_advance_esn) (struct xfrm_state *x); 914 }; 915 #endif 916 917 #if IS_ENABLED(CONFIG_TLS_DEVICE) 918 enum tls_offload_ctx_dir { 919 TLS_OFFLOAD_CTX_DIR_RX, 920 TLS_OFFLOAD_CTX_DIR_TX, 921 }; 922 923 struct tls_crypto_info; 924 struct tls_context; 925 926 struct tlsdev_ops { 927 int (*tls_dev_add)(struct net_device *netdev, struct sock *sk, 928 enum tls_offload_ctx_dir direction, 929 struct tls_crypto_info *crypto_info, 930 u32 start_offload_tcp_sn); 931 void (*tls_dev_del)(struct net_device *netdev, 932 struct tls_context *ctx, 933 enum tls_offload_ctx_dir direction); 934 void (*tls_dev_resync_rx)(struct net_device *netdev, 935 struct sock *sk, u32 seq, u64 rcd_sn); 936 }; 937 #endif 938 939 struct dev_ifalias { 940 struct rcu_head rcuhead; 941 char ifalias[]; 942 }; 943 944 /* 945 * This structure defines the management hooks for network devices. 946 * The following hooks can be defined; unless noted otherwise, they are 947 * optional and can be filled with a null pointer. 948 * 949 * int (*ndo_init)(struct net_device *dev); 950 * This function is called once when a network device is registered. 951 * The network device can use this for any late stage initialization 952 * or semantic validation. It can fail with an error code which will 953 * be propagated back to register_netdev. 954 * 955 * void (*ndo_uninit)(struct net_device *dev); 956 * This function is called when device is unregistered or when registration 957 * fails. It is not called if init fails. 958 * 959 * int (*ndo_open)(struct net_device *dev); 960 * This function is called when a network device transitions to the up 961 * state. 962 * 963 * int (*ndo_stop)(struct net_device *dev); 964 * This function is called when a network device transitions to the down 965 * state. 966 * 967 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb, 968 * struct net_device *dev); 969 * Called when a packet needs to be transmitted. 970 * Returns NETDEV_TX_OK. Can return NETDEV_TX_BUSY, but you should stop 971 * the queue before that can happen; it's for obsolete devices and weird 972 * corner cases, but the stack really does a non-trivial amount 973 * of useless work if you return NETDEV_TX_BUSY. 974 * Required; cannot be NULL. 975 * 976 * netdev_features_t (*ndo_features_check)(struct sk_buff *skb, 977 * struct net_device *dev 978 * netdev_features_t features); 979 * Called by core transmit path to determine if device is capable of 980 * performing offload operations on a given packet. This is to give 981 * the device an opportunity to implement any restrictions that cannot 982 * be otherwise expressed by feature flags. The check is called with 983 * the set of features that the stack has calculated and it returns 984 * those the driver believes to be appropriate. 985 * 986 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb, 987 * struct net_device *sb_dev, 988 * select_queue_fallback_t fallback); 989 * Called to decide which queue to use when device supports multiple 990 * transmit queues. 991 * 992 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags); 993 * This function is called to allow device receiver to make 994 * changes to configuration when multicast or promiscuous is enabled. 995 * 996 * void (*ndo_set_rx_mode)(struct net_device *dev); 997 * This function is called device changes address list filtering. 998 * If driver handles unicast address filtering, it should set 999 * IFF_UNICAST_FLT in its priv_flags. 1000 * 1001 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr); 1002 * This function is called when the Media Access Control address 1003 * needs to be changed. If this interface is not defined, the 1004 * MAC address can not be changed. 1005 * 1006 * int (*ndo_validate_addr)(struct net_device *dev); 1007 * Test if Media Access Control address is valid for the device. 1008 * 1009 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd); 1010 * Called when a user requests an ioctl which can't be handled by 1011 * the generic interface code. If not defined ioctls return 1012 * not supported error code. 1013 * 1014 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map); 1015 * Used to set network devices bus interface parameters. This interface 1016 * is retained for legacy reasons; new devices should use the bus 1017 * interface (PCI) for low level management. 1018 * 1019 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu); 1020 * Called when a user wants to change the Maximum Transfer Unit 1021 * of a device. 1022 * 1023 * void (*ndo_tx_timeout)(struct net_device *dev); 1024 * Callback used when the transmitter has not made any progress 1025 * for dev->watchdog ticks. 1026 * 1027 * void (*ndo_get_stats64)(struct net_device *dev, 1028 * struct rtnl_link_stats64 *storage); 1029 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev); 1030 * Called when a user wants to get the network device usage 1031 * statistics. Drivers must do one of the following: 1032 * 1. Define @ndo_get_stats64 to fill in a zero-initialised 1033 * rtnl_link_stats64 structure passed by the caller. 1034 * 2. Define @ndo_get_stats to update a net_device_stats structure 1035 * (which should normally be dev->stats) and return a pointer to 1036 * it. The structure may be changed asynchronously only if each 1037 * field is written atomically. 1038 * 3. Update dev->stats asynchronously and atomically, and define 1039 * neither operation. 1040 * 1041 * bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id) 1042 * Return true if this device supports offload stats of this attr_id. 1043 * 1044 * int (*ndo_get_offload_stats)(int attr_id, const struct net_device *dev, 1045 * void *attr_data) 1046 * Get statistics for offload operations by attr_id. Write it into the 1047 * attr_data pointer. 1048 * 1049 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16 vid); 1050 * If device supports VLAN filtering this function is called when a 1051 * VLAN id is registered. 1052 * 1053 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, __be16 proto, u16 vid); 1054 * If device supports VLAN filtering this function is called when a 1055 * VLAN id is unregistered. 1056 * 1057 * void (*ndo_poll_controller)(struct net_device *dev); 1058 * 1059 * SR-IOV management functions. 1060 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac); 1061 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan, 1062 * u8 qos, __be16 proto); 1063 * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate, 1064 * int max_tx_rate); 1065 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting); 1066 * int (*ndo_set_vf_trust)(struct net_device *dev, int vf, bool setting); 1067 * int (*ndo_get_vf_config)(struct net_device *dev, 1068 * int vf, struct ifla_vf_info *ivf); 1069 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state); 1070 * int (*ndo_set_vf_port)(struct net_device *dev, int vf, 1071 * struct nlattr *port[]); 1072 * 1073 * Enable or disable the VF ability to query its RSS Redirection Table and 1074 * Hash Key. This is needed since on some devices VF share this information 1075 * with PF and querying it may introduce a theoretical security risk. 1076 * int (*ndo_set_vf_rss_query_en)(struct net_device *dev, int vf, bool setting); 1077 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb); 1078 * int (*ndo_setup_tc)(struct net_device *dev, enum tc_setup_type type, 1079 * void *type_data); 1080 * Called to setup any 'tc' scheduler, classifier or action on @dev. 1081 * This is always called from the stack with the rtnl lock held and netif 1082 * tx queues stopped. This allows the netdevice to perform queue 1083 * management safely. 1084 * 1085 * Fiber Channel over Ethernet (FCoE) offload functions. 1086 * int (*ndo_fcoe_enable)(struct net_device *dev); 1087 * Called when the FCoE protocol stack wants to start using LLD for FCoE 1088 * so the underlying device can perform whatever needed configuration or 1089 * initialization to support acceleration of FCoE traffic. 1090 * 1091 * int (*ndo_fcoe_disable)(struct net_device *dev); 1092 * Called when the FCoE protocol stack wants to stop using LLD for FCoE 1093 * so the underlying device can perform whatever needed clean-ups to 1094 * stop supporting acceleration of FCoE traffic. 1095 * 1096 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid, 1097 * struct scatterlist *sgl, unsigned int sgc); 1098 * Called when the FCoE Initiator wants to initialize an I/O that 1099 * is a possible candidate for Direct Data Placement (DDP). The LLD can 1100 * perform necessary setup and returns 1 to indicate the device is set up 1101 * successfully to perform DDP on this I/O, otherwise this returns 0. 1102 * 1103 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid); 1104 * Called when the FCoE Initiator/Target is done with the DDPed I/O as 1105 * indicated by the FC exchange id 'xid', so the underlying device can 1106 * clean up and reuse resources for later DDP requests. 1107 * 1108 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid, 1109 * struct scatterlist *sgl, unsigned int sgc); 1110 * Called when the FCoE Target wants to initialize an I/O that 1111 * is a possible candidate for Direct Data Placement (DDP). The LLD can 1112 * perform necessary setup and returns 1 to indicate the device is set up 1113 * successfully to perform DDP on this I/O, otherwise this returns 0. 1114 * 1115 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev, 1116 * struct netdev_fcoe_hbainfo *hbainfo); 1117 * Called when the FCoE Protocol stack wants information on the underlying 1118 * device. This information is utilized by the FCoE protocol stack to 1119 * register attributes with Fiber Channel management service as per the 1120 * FC-GS Fabric Device Management Information(FDMI) specification. 1121 * 1122 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type); 1123 * Called when the underlying device wants to override default World Wide 1124 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own 1125 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE 1126 * protocol stack to use. 1127 * 1128 * RFS acceleration. 1129 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb, 1130 * u16 rxq_index, u32 flow_id); 1131 * Set hardware filter for RFS. rxq_index is the target queue index; 1132 * flow_id is a flow ID to be passed to rps_may_expire_flow() later. 1133 * Return the filter ID on success, or a negative error code. 1134 * 1135 * Slave management functions (for bridge, bonding, etc). 1136 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev); 1137 * Called to make another netdev an underling. 1138 * 1139 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev); 1140 * Called to release previously enslaved netdev. 1141 * 1142 * Feature/offload setting functions. 1143 * netdev_features_t (*ndo_fix_features)(struct net_device *dev, 1144 * netdev_features_t features); 1145 * Adjusts the requested feature flags according to device-specific 1146 * constraints, and returns the resulting flags. Must not modify 1147 * the device state. 1148 * 1149 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features); 1150 * Called to update device configuration to new features. Passed 1151 * feature set might be less than what was returned by ndo_fix_features()). 1152 * Must return >0 or -errno if it changed dev->features itself. 1153 * 1154 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[], 1155 * struct net_device *dev, 1156 * const unsigned char *addr, u16 vid, u16 flags, 1157 * struct netlink_ext_ack *extack); 1158 * Adds an FDB entry to dev for addr. 1159 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[], 1160 * struct net_device *dev, 1161 * const unsigned char *addr, u16 vid) 1162 * Deletes the FDB entry from dev coresponding to addr. 1163 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb, 1164 * struct net_device *dev, struct net_device *filter_dev, 1165 * int *idx) 1166 * Used to add FDB entries to dump requests. Implementers should add 1167 * entries to skb and update idx with the number of entries. 1168 * 1169 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh, 1170 * u16 flags, struct netlink_ext_ack *extack) 1171 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq, 1172 * struct net_device *dev, u32 filter_mask, 1173 * int nlflags) 1174 * int (*ndo_bridge_dellink)(struct net_device *dev, struct nlmsghdr *nlh, 1175 * u16 flags); 1176 * 1177 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier); 1178 * Called to change device carrier. Soft-devices (like dummy, team, etc) 1179 * which do not represent real hardware may define this to allow their 1180 * userspace components to manage their virtual carrier state. Devices 1181 * that determine carrier state from physical hardware properties (eg 1182 * network cables) or protocol-dependent mechanisms (eg 1183 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function. 1184 * 1185 * int (*ndo_get_phys_port_id)(struct net_device *dev, 1186 * struct netdev_phys_item_id *ppid); 1187 * Called to get ID of physical port of this device. If driver does 1188 * not implement this, it is assumed that the hw is not able to have 1189 * multiple net devices on single physical port. 1190 * 1191 * int (*ndo_get_port_parent_id)(struct net_device *dev, 1192 * struct netdev_phys_item_id *ppid) 1193 * Called to get the parent ID of the physical port of this device. 1194 * 1195 * void (*ndo_udp_tunnel_add)(struct net_device *dev, 1196 * struct udp_tunnel_info *ti); 1197 * Called by UDP tunnel to notify a driver about the UDP port and socket 1198 * address family that a UDP tunnel is listnening to. It is called only 1199 * when a new port starts listening. The operation is protected by the 1200 * RTNL. 1201 * 1202 * void (*ndo_udp_tunnel_del)(struct net_device *dev, 1203 * struct udp_tunnel_info *ti); 1204 * Called by UDP tunnel to notify the driver about a UDP port and socket 1205 * address family that the UDP tunnel is not listening to anymore. The 1206 * operation is protected by the RTNL. 1207 * 1208 * void* (*ndo_dfwd_add_station)(struct net_device *pdev, 1209 * struct net_device *dev) 1210 * Called by upper layer devices to accelerate switching or other 1211 * station functionality into hardware. 'pdev is the lowerdev 1212 * to use for the offload and 'dev' is the net device that will 1213 * back the offload. Returns a pointer to the private structure 1214 * the upper layer will maintain. 1215 * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv) 1216 * Called by upper layer device to delete the station created 1217 * by 'ndo_dfwd_add_station'. 'pdev' is the net device backing 1218 * the station and priv is the structure returned by the add 1219 * operation. 1220 * int (*ndo_set_tx_maxrate)(struct net_device *dev, 1221 * int queue_index, u32 maxrate); 1222 * Called when a user wants to set a max-rate limitation of specific 1223 * TX queue. 1224 * int (*ndo_get_iflink)(const struct net_device *dev); 1225 * Called to get the iflink value of this device. 1226 * void (*ndo_change_proto_down)(struct net_device *dev, 1227 * bool proto_down); 1228 * This function is used to pass protocol port error state information 1229 * to the switch driver. The switch driver can react to the proto_down 1230 * by doing a phys down on the associated switch port. 1231 * int (*ndo_fill_metadata_dst)(struct net_device *dev, struct sk_buff *skb); 1232 * This function is used to get egress tunnel information for given skb. 1233 * This is useful for retrieving outer tunnel header parameters while 1234 * sampling packet. 1235 * void (*ndo_set_rx_headroom)(struct net_device *dev, int needed_headroom); 1236 * This function is used to specify the headroom that the skb must 1237 * consider when allocation skb during packet reception. Setting 1238 * appropriate rx headroom value allows avoiding skb head copy on 1239 * forward. Setting a negative value resets the rx headroom to the 1240 * default value. 1241 * int (*ndo_bpf)(struct net_device *dev, struct netdev_bpf *bpf); 1242 * This function is used to set or query state related to XDP on the 1243 * netdevice and manage BPF offload. See definition of 1244 * enum bpf_netdev_command for details. 1245 * int (*ndo_xdp_xmit)(struct net_device *dev, int n, struct xdp_frame **xdp, 1246 * u32 flags); 1247 * This function is used to submit @n XDP packets for transmit on a 1248 * netdevice. Returns number of frames successfully transmitted, frames 1249 * that got dropped are freed/returned via xdp_return_frame(). 1250 * Returns negative number, means general error invoking ndo, meaning 1251 * no frames were xmit'ed and core-caller will free all frames. 1252 */ 1253 struct net_device_ops { 1254 int (*ndo_init)(struct net_device *dev); 1255 void (*ndo_uninit)(struct net_device *dev); 1256 int (*ndo_open)(struct net_device *dev); 1257 int (*ndo_stop)(struct net_device *dev); 1258 netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb, 1259 struct net_device *dev); 1260 netdev_features_t (*ndo_features_check)(struct sk_buff *skb, 1261 struct net_device *dev, 1262 netdev_features_t features); 1263 u16 (*ndo_select_queue)(struct net_device *dev, 1264 struct sk_buff *skb, 1265 struct net_device *sb_dev, 1266 select_queue_fallback_t fallback); 1267 void (*ndo_change_rx_flags)(struct net_device *dev, 1268 int flags); 1269 void (*ndo_set_rx_mode)(struct net_device *dev); 1270 int (*ndo_set_mac_address)(struct net_device *dev, 1271 void *addr); 1272 int (*ndo_validate_addr)(struct net_device *dev); 1273 int (*ndo_do_ioctl)(struct net_device *dev, 1274 struct ifreq *ifr, int cmd); 1275 int (*ndo_set_config)(struct net_device *dev, 1276 struct ifmap *map); 1277 int (*ndo_change_mtu)(struct net_device *dev, 1278 int new_mtu); 1279 int (*ndo_neigh_setup)(struct net_device *dev, 1280 struct neigh_parms *); 1281 void (*ndo_tx_timeout) (struct net_device *dev); 1282 1283 void (*ndo_get_stats64)(struct net_device *dev, 1284 struct rtnl_link_stats64 *storage); 1285 bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id); 1286 int (*ndo_get_offload_stats)(int attr_id, 1287 const struct net_device *dev, 1288 void *attr_data); 1289 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev); 1290 1291 int (*ndo_vlan_rx_add_vid)(struct net_device *dev, 1292 __be16 proto, u16 vid); 1293 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, 1294 __be16 proto, u16 vid); 1295 #ifdef CONFIG_NET_POLL_CONTROLLER 1296 void (*ndo_poll_controller)(struct net_device *dev); 1297 int (*ndo_netpoll_setup)(struct net_device *dev, 1298 struct netpoll_info *info); 1299 void (*ndo_netpoll_cleanup)(struct net_device *dev); 1300 #endif 1301 int (*ndo_set_vf_mac)(struct net_device *dev, 1302 int queue, u8 *mac); 1303 int (*ndo_set_vf_vlan)(struct net_device *dev, 1304 int queue, u16 vlan, 1305 u8 qos, __be16 proto); 1306 int (*ndo_set_vf_rate)(struct net_device *dev, 1307 int vf, int min_tx_rate, 1308 int max_tx_rate); 1309 int (*ndo_set_vf_spoofchk)(struct net_device *dev, 1310 int vf, bool setting); 1311 int (*ndo_set_vf_trust)(struct net_device *dev, 1312 int vf, bool setting); 1313 int (*ndo_get_vf_config)(struct net_device *dev, 1314 int vf, 1315 struct ifla_vf_info *ivf); 1316 int (*ndo_set_vf_link_state)(struct net_device *dev, 1317 int vf, int link_state); 1318 int (*ndo_get_vf_stats)(struct net_device *dev, 1319 int vf, 1320 struct ifla_vf_stats 1321 *vf_stats); 1322 int (*ndo_set_vf_port)(struct net_device *dev, 1323 int vf, 1324 struct nlattr *port[]); 1325 int (*ndo_get_vf_port)(struct net_device *dev, 1326 int vf, struct sk_buff *skb); 1327 int (*ndo_set_vf_guid)(struct net_device *dev, 1328 int vf, u64 guid, 1329 int guid_type); 1330 int (*ndo_set_vf_rss_query_en)( 1331 struct net_device *dev, 1332 int vf, bool setting); 1333 int (*ndo_setup_tc)(struct net_device *dev, 1334 enum tc_setup_type type, 1335 void *type_data); 1336 #if IS_ENABLED(CONFIG_FCOE) 1337 int (*ndo_fcoe_enable)(struct net_device *dev); 1338 int (*ndo_fcoe_disable)(struct net_device *dev); 1339 int (*ndo_fcoe_ddp_setup)(struct net_device *dev, 1340 u16 xid, 1341 struct scatterlist *sgl, 1342 unsigned int sgc); 1343 int (*ndo_fcoe_ddp_done)(struct net_device *dev, 1344 u16 xid); 1345 int (*ndo_fcoe_ddp_target)(struct net_device *dev, 1346 u16 xid, 1347 struct scatterlist *sgl, 1348 unsigned int sgc); 1349 int (*ndo_fcoe_get_hbainfo)(struct net_device *dev, 1350 struct netdev_fcoe_hbainfo *hbainfo); 1351 #endif 1352 1353 #if IS_ENABLED(CONFIG_LIBFCOE) 1354 #define NETDEV_FCOE_WWNN 0 1355 #define NETDEV_FCOE_WWPN 1 1356 int (*ndo_fcoe_get_wwn)(struct net_device *dev, 1357 u64 *wwn, int type); 1358 #endif 1359 1360 #ifdef CONFIG_RFS_ACCEL 1361 int (*ndo_rx_flow_steer)(struct net_device *dev, 1362 const struct sk_buff *skb, 1363 u16 rxq_index, 1364 u32 flow_id); 1365 #endif 1366 int (*ndo_add_slave)(struct net_device *dev, 1367 struct net_device *slave_dev, 1368 struct netlink_ext_ack *extack); 1369 int (*ndo_del_slave)(struct net_device *dev, 1370 struct net_device *slave_dev); 1371 netdev_features_t (*ndo_fix_features)(struct net_device *dev, 1372 netdev_features_t features); 1373 int (*ndo_set_features)(struct net_device *dev, 1374 netdev_features_t features); 1375 int (*ndo_neigh_construct)(struct net_device *dev, 1376 struct neighbour *n); 1377 void (*ndo_neigh_destroy)(struct net_device *dev, 1378 struct neighbour *n); 1379 1380 int (*ndo_fdb_add)(struct ndmsg *ndm, 1381 struct nlattr *tb[], 1382 struct net_device *dev, 1383 const unsigned char *addr, 1384 u16 vid, 1385 u16 flags, 1386 struct netlink_ext_ack *extack); 1387 int (*ndo_fdb_del)(struct ndmsg *ndm, 1388 struct nlattr *tb[], 1389 struct net_device *dev, 1390 const unsigned char *addr, 1391 u16 vid); 1392 int (*ndo_fdb_dump)(struct sk_buff *skb, 1393 struct netlink_callback *cb, 1394 struct net_device *dev, 1395 struct net_device *filter_dev, 1396 int *idx); 1397 int (*ndo_fdb_get)(struct sk_buff *skb, 1398 struct nlattr *tb[], 1399 struct net_device *dev, 1400 const unsigned char *addr, 1401 u16 vid, u32 portid, u32 seq, 1402 struct netlink_ext_ack *extack); 1403 int (*ndo_bridge_setlink)(struct net_device *dev, 1404 struct nlmsghdr *nlh, 1405 u16 flags, 1406 struct netlink_ext_ack *extack); 1407 int (*ndo_bridge_getlink)(struct sk_buff *skb, 1408 u32 pid, u32 seq, 1409 struct net_device *dev, 1410 u32 filter_mask, 1411 int nlflags); 1412 int (*ndo_bridge_dellink)(struct net_device *dev, 1413 struct nlmsghdr *nlh, 1414 u16 flags); 1415 int (*ndo_change_carrier)(struct net_device *dev, 1416 bool new_carrier); 1417 int (*ndo_get_phys_port_id)(struct net_device *dev, 1418 struct netdev_phys_item_id *ppid); 1419 int (*ndo_get_port_parent_id)(struct net_device *dev, 1420 struct netdev_phys_item_id *ppid); 1421 int (*ndo_get_phys_port_name)(struct net_device *dev, 1422 char *name, size_t len); 1423 void (*ndo_udp_tunnel_add)(struct net_device *dev, 1424 struct udp_tunnel_info *ti); 1425 void (*ndo_udp_tunnel_del)(struct net_device *dev, 1426 struct udp_tunnel_info *ti); 1427 void* (*ndo_dfwd_add_station)(struct net_device *pdev, 1428 struct net_device *dev); 1429 void (*ndo_dfwd_del_station)(struct net_device *pdev, 1430 void *priv); 1431 1432 int (*ndo_get_lock_subclass)(struct net_device *dev); 1433 int (*ndo_set_tx_maxrate)(struct net_device *dev, 1434 int queue_index, 1435 u32 maxrate); 1436 int (*ndo_get_iflink)(const struct net_device *dev); 1437 int (*ndo_change_proto_down)(struct net_device *dev, 1438 bool proto_down); 1439 int (*ndo_fill_metadata_dst)(struct net_device *dev, 1440 struct sk_buff *skb); 1441 void (*ndo_set_rx_headroom)(struct net_device *dev, 1442 int needed_headroom); 1443 int (*ndo_bpf)(struct net_device *dev, 1444 struct netdev_bpf *bpf); 1445 int (*ndo_xdp_xmit)(struct net_device *dev, int n, 1446 struct xdp_frame **xdp, 1447 u32 flags); 1448 int (*ndo_xsk_async_xmit)(struct net_device *dev, 1449 u32 queue_id); 1450 }; 1451 1452 /** 1453 * enum net_device_priv_flags - &struct net_device priv_flags 1454 * 1455 * These are the &struct net_device, they are only set internally 1456 * by drivers and used in the kernel. These flags are invisible to 1457 * userspace; this means that the order of these flags can change 1458 * during any kernel release. 1459 * 1460 * You should have a pretty good reason to be extending these flags. 1461 * 1462 * @IFF_802_1Q_VLAN: 802.1Q VLAN device 1463 * @IFF_EBRIDGE: Ethernet bridging device 1464 * @IFF_BONDING: bonding master or slave 1465 * @IFF_ISATAP: ISATAP interface (RFC4214) 1466 * @IFF_WAN_HDLC: WAN HDLC device 1467 * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to 1468 * release skb->dst 1469 * @IFF_DONT_BRIDGE: disallow bridging this ether dev 1470 * @IFF_DISABLE_NETPOLL: disable netpoll at run-time 1471 * @IFF_MACVLAN_PORT: device used as macvlan port 1472 * @IFF_BRIDGE_PORT: device used as bridge port 1473 * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port 1474 * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit 1475 * @IFF_UNICAST_FLT: Supports unicast filtering 1476 * @IFF_TEAM_PORT: device used as team port 1477 * @IFF_SUPP_NOFCS: device supports sending custom FCS 1478 * @IFF_LIVE_ADDR_CHANGE: device supports hardware address 1479 * change when it's running 1480 * @IFF_MACVLAN: Macvlan device 1481 * @IFF_XMIT_DST_RELEASE_PERM: IFF_XMIT_DST_RELEASE not taking into account 1482 * underlying stacked devices 1483 * @IFF_L3MDEV_MASTER: device is an L3 master device 1484 * @IFF_NO_QUEUE: device can run without qdisc attached 1485 * @IFF_OPENVSWITCH: device is a Open vSwitch master 1486 * @IFF_L3MDEV_SLAVE: device is enslaved to an L3 master device 1487 * @IFF_TEAM: device is a team device 1488 * @IFF_RXFH_CONFIGURED: device has had Rx Flow indirection table configured 1489 * @IFF_PHONY_HEADROOM: the headroom value is controlled by an external 1490 * entity (i.e. the master device for bridged veth) 1491 * @IFF_MACSEC: device is a MACsec device 1492 * @IFF_NO_RX_HANDLER: device doesn't support the rx_handler hook 1493 * @IFF_FAILOVER: device is a failover master device 1494 * @IFF_FAILOVER_SLAVE: device is lower dev of a failover master device 1495 * @IFF_L3MDEV_RX_HANDLER: only invoke the rx handler of L3 master device 1496 */ 1497 enum netdev_priv_flags { 1498 IFF_802_1Q_VLAN = 1<<0, 1499 IFF_EBRIDGE = 1<<1, 1500 IFF_BONDING = 1<<2, 1501 IFF_ISATAP = 1<<3, 1502 IFF_WAN_HDLC = 1<<4, 1503 IFF_XMIT_DST_RELEASE = 1<<5, 1504 IFF_DONT_BRIDGE = 1<<6, 1505 IFF_DISABLE_NETPOLL = 1<<7, 1506 IFF_MACVLAN_PORT = 1<<8, 1507 IFF_BRIDGE_PORT = 1<<9, 1508 IFF_OVS_DATAPATH = 1<<10, 1509 IFF_TX_SKB_SHARING = 1<<11, 1510 IFF_UNICAST_FLT = 1<<12, 1511 IFF_TEAM_PORT = 1<<13, 1512 IFF_SUPP_NOFCS = 1<<14, 1513 IFF_LIVE_ADDR_CHANGE = 1<<15, 1514 IFF_MACVLAN = 1<<16, 1515 IFF_XMIT_DST_RELEASE_PERM = 1<<17, 1516 IFF_L3MDEV_MASTER = 1<<18, 1517 IFF_NO_QUEUE = 1<<19, 1518 IFF_OPENVSWITCH = 1<<20, 1519 IFF_L3MDEV_SLAVE = 1<<21, 1520 IFF_TEAM = 1<<22, 1521 IFF_RXFH_CONFIGURED = 1<<23, 1522 IFF_PHONY_HEADROOM = 1<<24, 1523 IFF_MACSEC = 1<<25, 1524 IFF_NO_RX_HANDLER = 1<<26, 1525 IFF_FAILOVER = 1<<27, 1526 IFF_FAILOVER_SLAVE = 1<<28, 1527 IFF_L3MDEV_RX_HANDLER = 1<<29, 1528 }; 1529 1530 #define IFF_802_1Q_VLAN IFF_802_1Q_VLAN 1531 #define IFF_EBRIDGE IFF_EBRIDGE 1532 #define IFF_BONDING IFF_BONDING 1533 #define IFF_ISATAP IFF_ISATAP 1534 #define IFF_WAN_HDLC IFF_WAN_HDLC 1535 #define IFF_XMIT_DST_RELEASE IFF_XMIT_DST_RELEASE 1536 #define IFF_DONT_BRIDGE IFF_DONT_BRIDGE 1537 #define IFF_DISABLE_NETPOLL IFF_DISABLE_NETPOLL 1538 #define IFF_MACVLAN_PORT IFF_MACVLAN_PORT 1539 #define IFF_BRIDGE_PORT IFF_BRIDGE_PORT 1540 #define IFF_OVS_DATAPATH IFF_OVS_DATAPATH 1541 #define IFF_TX_SKB_SHARING IFF_TX_SKB_SHARING 1542 #define IFF_UNICAST_FLT IFF_UNICAST_FLT 1543 #define IFF_TEAM_PORT IFF_TEAM_PORT 1544 #define IFF_SUPP_NOFCS IFF_SUPP_NOFCS 1545 #define IFF_LIVE_ADDR_CHANGE IFF_LIVE_ADDR_CHANGE 1546 #define IFF_MACVLAN IFF_MACVLAN 1547 #define IFF_XMIT_DST_RELEASE_PERM IFF_XMIT_DST_RELEASE_PERM 1548 #define IFF_L3MDEV_MASTER IFF_L3MDEV_MASTER 1549 #define IFF_NO_QUEUE IFF_NO_QUEUE 1550 #define IFF_OPENVSWITCH IFF_OPENVSWITCH 1551 #define IFF_L3MDEV_SLAVE IFF_L3MDEV_SLAVE 1552 #define IFF_TEAM IFF_TEAM 1553 #define IFF_RXFH_CONFIGURED IFF_RXFH_CONFIGURED 1554 #define IFF_MACSEC IFF_MACSEC 1555 #define IFF_NO_RX_HANDLER IFF_NO_RX_HANDLER 1556 #define IFF_FAILOVER IFF_FAILOVER 1557 #define IFF_FAILOVER_SLAVE IFF_FAILOVER_SLAVE 1558 #define IFF_L3MDEV_RX_HANDLER IFF_L3MDEV_RX_HANDLER 1559 1560 /** 1561 * struct net_device - The DEVICE structure. 1562 * 1563 * Actually, this whole structure is a big mistake. It mixes I/O 1564 * data with strictly "high-level" data, and it has to know about 1565 * almost every data structure used in the INET module. 1566 * 1567 * @name: This is the first field of the "visible" part of this structure 1568 * (i.e. as seen by users in the "Space.c" file). It is the name 1569 * of the interface. 1570 * 1571 * @name_hlist: Device name hash chain, please keep it close to name[] 1572 * @ifalias: SNMP alias 1573 * @mem_end: Shared memory end 1574 * @mem_start: Shared memory start 1575 * @base_addr: Device I/O address 1576 * @irq: Device IRQ number 1577 * 1578 * @state: Generic network queuing layer state, see netdev_state_t 1579 * @dev_list: The global list of network devices 1580 * @napi_list: List entry used for polling NAPI devices 1581 * @unreg_list: List entry when we are unregistering the 1582 * device; see the function unregister_netdev 1583 * @close_list: List entry used when we are closing the device 1584 * @ptype_all: Device-specific packet handlers for all protocols 1585 * @ptype_specific: Device-specific, protocol-specific packet handlers 1586 * 1587 * @adj_list: Directly linked devices, like slaves for bonding 1588 * @features: Currently active device features 1589 * @hw_features: User-changeable features 1590 * 1591 * @wanted_features: User-requested features 1592 * @vlan_features: Mask of features inheritable by VLAN devices 1593 * 1594 * @hw_enc_features: Mask of features inherited by encapsulating devices 1595 * This field indicates what encapsulation 1596 * offloads the hardware is capable of doing, 1597 * and drivers will need to set them appropriately. 1598 * 1599 * @mpls_features: Mask of features inheritable by MPLS 1600 * 1601 * @ifindex: interface index 1602 * @group: The group the device belongs to 1603 * 1604 * @stats: Statistics struct, which was left as a legacy, use 1605 * rtnl_link_stats64 instead 1606 * 1607 * @rx_dropped: Dropped packets by core network, 1608 * do not use this in drivers 1609 * @tx_dropped: Dropped packets by core network, 1610 * do not use this in drivers 1611 * @rx_nohandler: nohandler dropped packets by core network on 1612 * inactive devices, do not use this in drivers 1613 * @carrier_up_count: Number of times the carrier has been up 1614 * @carrier_down_count: Number of times the carrier has been down 1615 * 1616 * @wireless_handlers: List of functions to handle Wireless Extensions, 1617 * instead of ioctl, 1618 * see <net/iw_handler.h> for details. 1619 * @wireless_data: Instance data managed by the core of wireless extensions 1620 * 1621 * @netdev_ops: Includes several pointers to callbacks, 1622 * if one wants to override the ndo_*() functions 1623 * @ethtool_ops: Management operations 1624 * @ndisc_ops: Includes callbacks for different IPv6 neighbour 1625 * discovery handling. Necessary for e.g. 6LoWPAN. 1626 * @header_ops: Includes callbacks for creating,parsing,caching,etc 1627 * of Layer 2 headers. 1628 * 1629 * @flags: Interface flags (a la BSD) 1630 * @priv_flags: Like 'flags' but invisible to userspace, 1631 * see if.h for the definitions 1632 * @gflags: Global flags ( kept as legacy ) 1633 * @padded: How much padding added by alloc_netdev() 1634 * @operstate: RFC2863 operstate 1635 * @link_mode: Mapping policy to operstate 1636 * @if_port: Selectable AUI, TP, ... 1637 * @dma: DMA channel 1638 * @mtu: Interface MTU value 1639 * @min_mtu: Interface Minimum MTU value 1640 * @max_mtu: Interface Maximum MTU value 1641 * @type: Interface hardware type 1642 * @hard_header_len: Maximum hardware header length. 1643 * @min_header_len: Minimum hardware header length 1644 * 1645 * @needed_headroom: Extra headroom the hardware may need, but not in all 1646 * cases can this be guaranteed 1647 * @needed_tailroom: Extra tailroom the hardware may need, but not in all 1648 * cases can this be guaranteed. Some cases also use 1649 * LL_MAX_HEADER instead to allocate the skb 1650 * 1651 * interface address info: 1652 * 1653 * @perm_addr: Permanent hw address 1654 * @addr_assign_type: Hw address assignment type 1655 * @addr_len: Hardware address length 1656 * @neigh_priv_len: Used in neigh_alloc() 1657 * @dev_id: Used to differentiate devices that share 1658 * the same link layer address 1659 * @dev_port: Used to differentiate devices that share 1660 * the same function 1661 * @addr_list_lock: XXX: need comments on this one 1662 * @uc_promisc: Counter that indicates promiscuous mode 1663 * has been enabled due to the need to listen to 1664 * additional unicast addresses in a device that 1665 * does not implement ndo_set_rx_mode() 1666 * @uc: unicast mac addresses 1667 * @mc: multicast mac addresses 1668 * @dev_addrs: list of device hw addresses 1669 * @queues_kset: Group of all Kobjects in the Tx and RX queues 1670 * @promiscuity: Number of times the NIC is told to work in 1671 * promiscuous mode; if it becomes 0 the NIC will 1672 * exit promiscuous mode 1673 * @allmulti: Counter, enables or disables allmulticast mode 1674 * 1675 * @vlan_info: VLAN info 1676 * @dsa_ptr: dsa specific data 1677 * @tipc_ptr: TIPC specific data 1678 * @atalk_ptr: AppleTalk link 1679 * @ip_ptr: IPv4 specific data 1680 * @dn_ptr: DECnet specific data 1681 * @ip6_ptr: IPv6 specific data 1682 * @ax25_ptr: AX.25 specific data 1683 * @ieee80211_ptr: IEEE 802.11 specific data, assign before registering 1684 * 1685 * @dev_addr: Hw address (before bcast, 1686 * because most packets are unicast) 1687 * 1688 * @_rx: Array of RX queues 1689 * @num_rx_queues: Number of RX queues 1690 * allocated at register_netdev() time 1691 * @real_num_rx_queues: Number of RX queues currently active in device 1692 * 1693 * @rx_handler: handler for received packets 1694 * @rx_handler_data: XXX: need comments on this one 1695 * @miniq_ingress: ingress/clsact qdisc specific data for 1696 * ingress processing 1697 * @ingress_queue: XXX: need comments on this one 1698 * @broadcast: hw bcast address 1699 * 1700 * @rx_cpu_rmap: CPU reverse-mapping for RX completion interrupts, 1701 * indexed by RX queue number. Assigned by driver. 1702 * This must only be set if the ndo_rx_flow_steer 1703 * operation is defined 1704 * @index_hlist: Device index hash chain 1705 * 1706 * @_tx: Array of TX queues 1707 * @num_tx_queues: Number of TX queues allocated at alloc_netdev_mq() time 1708 * @real_num_tx_queues: Number of TX queues currently active in device 1709 * @qdisc: Root qdisc from userspace point of view 1710 * @tx_queue_len: Max frames per queue allowed 1711 * @tx_global_lock: XXX: need comments on this one 1712 * 1713 * @xps_maps: XXX: need comments on this one 1714 * @miniq_egress: clsact qdisc specific data for 1715 * egress processing 1716 * @watchdog_timeo: Represents the timeout that is used by 1717 * the watchdog (see dev_watchdog()) 1718 * @watchdog_timer: List of timers 1719 * 1720 * @pcpu_refcnt: Number of references to this device 1721 * @todo_list: Delayed register/unregister 1722 * @link_watch_list: XXX: need comments on this one 1723 * 1724 * @reg_state: Register/unregister state machine 1725 * @dismantle: Device is going to be freed 1726 * @rtnl_link_state: This enum represents the phases of creating 1727 * a new link 1728 * 1729 * @needs_free_netdev: Should unregister perform free_netdev? 1730 * @priv_destructor: Called from unregister 1731 * @npinfo: XXX: need comments on this one 1732 * @nd_net: Network namespace this network device is inside 1733 * 1734 * @ml_priv: Mid-layer private 1735 * @lstats: Loopback statistics 1736 * @tstats: Tunnel statistics 1737 * @dstats: Dummy statistics 1738 * @vstats: Virtual ethernet statistics 1739 * 1740 * @garp_port: GARP 1741 * @mrp_port: MRP 1742 * 1743 * @dev: Class/net/name entry 1744 * @sysfs_groups: Space for optional device, statistics and wireless 1745 * sysfs groups 1746 * 1747 * @sysfs_rx_queue_group: Space for optional per-rx queue attributes 1748 * @rtnl_link_ops: Rtnl_link_ops 1749 * 1750 * @gso_max_size: Maximum size of generic segmentation offload 1751 * @gso_max_segs: Maximum number of segments that can be passed to the 1752 * NIC for GSO 1753 * 1754 * @dcbnl_ops: Data Center Bridging netlink ops 1755 * @num_tc: Number of traffic classes in the net device 1756 * @tc_to_txq: XXX: need comments on this one 1757 * @prio_tc_map: XXX: need comments on this one 1758 * 1759 * @fcoe_ddp_xid: Max exchange id for FCoE LRO by ddp 1760 * 1761 * @priomap: XXX: need comments on this one 1762 * @phydev: Physical device may attach itself 1763 * for hardware timestamping 1764 * @sfp_bus: attached &struct sfp_bus structure. 1765 * 1766 * @qdisc_tx_busylock: lockdep class annotating Qdisc->busylock spinlock 1767 * @qdisc_running_key: lockdep class annotating Qdisc->running seqcount 1768 * 1769 * @proto_down: protocol port state information can be sent to the 1770 * switch driver and used to set the phys state of the 1771 * switch port. 1772 * 1773 * @wol_enabled: Wake-on-LAN is enabled 1774 * 1775 * FIXME: cleanup struct net_device such that network protocol info 1776 * moves out. 1777 */ 1778 1779 struct net_device { 1780 char name[IFNAMSIZ]; 1781 struct hlist_node name_hlist; 1782 struct dev_ifalias __rcu *ifalias; 1783 /* 1784 * I/O specific fields 1785 * FIXME: Merge these and struct ifmap into one 1786 */ 1787 unsigned long mem_end; 1788 unsigned long mem_start; 1789 unsigned long base_addr; 1790 int irq; 1791 1792 /* 1793 * Some hardware also needs these fields (state,dev_list, 1794 * napi_list,unreg_list,close_list) but they are not 1795 * part of the usual set specified in Space.c. 1796 */ 1797 1798 unsigned long state; 1799 1800 struct list_head dev_list; 1801 struct list_head napi_list; 1802 struct list_head unreg_list; 1803 struct list_head close_list; 1804 struct list_head ptype_all; 1805 struct list_head ptype_specific; 1806 1807 struct { 1808 struct list_head upper; 1809 struct list_head lower; 1810 } adj_list; 1811 1812 netdev_features_t features; 1813 netdev_features_t hw_features; 1814 netdev_features_t wanted_features; 1815 netdev_features_t vlan_features; 1816 netdev_features_t hw_enc_features; 1817 netdev_features_t mpls_features; 1818 netdev_features_t gso_partial_features; 1819 1820 int ifindex; 1821 int group; 1822 1823 struct net_device_stats stats; 1824 1825 atomic_long_t rx_dropped; 1826 atomic_long_t tx_dropped; 1827 atomic_long_t rx_nohandler; 1828 1829 /* Stats to monitor link on/off, flapping */ 1830 atomic_t carrier_up_count; 1831 atomic_t carrier_down_count; 1832 1833 #ifdef CONFIG_WIRELESS_EXT 1834 const struct iw_handler_def *wireless_handlers; 1835 struct iw_public_data *wireless_data; 1836 #endif 1837 const struct net_device_ops *netdev_ops; 1838 const struct ethtool_ops *ethtool_ops; 1839 #ifdef CONFIG_NET_SWITCHDEV 1840 const struct switchdev_ops *switchdev_ops; 1841 #endif 1842 #ifdef CONFIG_NET_L3_MASTER_DEV 1843 const struct l3mdev_ops *l3mdev_ops; 1844 #endif 1845 #if IS_ENABLED(CONFIG_IPV6) 1846 const struct ndisc_ops *ndisc_ops; 1847 #endif 1848 1849 #ifdef CONFIG_XFRM_OFFLOAD 1850 const struct xfrmdev_ops *xfrmdev_ops; 1851 #endif 1852 1853 #if IS_ENABLED(CONFIG_TLS_DEVICE) 1854 const struct tlsdev_ops *tlsdev_ops; 1855 #endif 1856 1857 const struct header_ops *header_ops; 1858 1859 unsigned int flags; 1860 unsigned int priv_flags; 1861 1862 unsigned short gflags; 1863 unsigned short padded; 1864 1865 unsigned char operstate; 1866 unsigned char link_mode; 1867 1868 unsigned char if_port; 1869 unsigned char dma; 1870 1871 unsigned int mtu; 1872 unsigned int min_mtu; 1873 unsigned int max_mtu; 1874 unsigned short type; 1875 unsigned short hard_header_len; 1876 unsigned char min_header_len; 1877 1878 unsigned short needed_headroom; 1879 unsigned short needed_tailroom; 1880 1881 /* Interface address info. */ 1882 unsigned char perm_addr[MAX_ADDR_LEN]; 1883 unsigned char addr_assign_type; 1884 unsigned char addr_len; 1885 unsigned short neigh_priv_len; 1886 unsigned short dev_id; 1887 unsigned short dev_port; 1888 spinlock_t addr_list_lock; 1889 unsigned char name_assign_type; 1890 bool uc_promisc; 1891 struct netdev_hw_addr_list uc; 1892 struct netdev_hw_addr_list mc; 1893 struct netdev_hw_addr_list dev_addrs; 1894 1895 #ifdef CONFIG_SYSFS 1896 struct kset *queues_kset; 1897 #endif 1898 unsigned int promiscuity; 1899 unsigned int allmulti; 1900 1901 1902 /* Protocol-specific pointers */ 1903 1904 #if IS_ENABLED(CONFIG_VLAN_8021Q) 1905 struct vlan_info __rcu *vlan_info; 1906 #endif 1907 #if IS_ENABLED(CONFIG_NET_DSA) 1908 struct dsa_port *dsa_ptr; 1909 #endif 1910 #if IS_ENABLED(CONFIG_TIPC) 1911 struct tipc_bearer __rcu *tipc_ptr; 1912 #endif 1913 #if IS_ENABLED(CONFIG_IRDA) || IS_ENABLED(CONFIG_ATALK) 1914 void *atalk_ptr; 1915 #endif 1916 struct in_device __rcu *ip_ptr; 1917 #if IS_ENABLED(CONFIG_DECNET) 1918 struct dn_dev __rcu *dn_ptr; 1919 #endif 1920 struct inet6_dev __rcu *ip6_ptr; 1921 #if IS_ENABLED(CONFIG_AX25) 1922 void *ax25_ptr; 1923 #endif 1924 struct wireless_dev *ieee80211_ptr; 1925 struct wpan_dev *ieee802154_ptr; 1926 #if IS_ENABLED(CONFIG_MPLS_ROUTING) 1927 struct mpls_dev __rcu *mpls_ptr; 1928 #endif 1929 1930 /* 1931 * Cache lines mostly used on receive path (including eth_type_trans()) 1932 */ 1933 /* Interface address info used in eth_type_trans() */ 1934 unsigned char *dev_addr; 1935 1936 struct netdev_rx_queue *_rx; 1937 unsigned int num_rx_queues; 1938 unsigned int real_num_rx_queues; 1939 1940 struct bpf_prog __rcu *xdp_prog; 1941 unsigned long gro_flush_timeout; 1942 rx_handler_func_t __rcu *rx_handler; 1943 void __rcu *rx_handler_data; 1944 1945 #ifdef CONFIG_NET_CLS_ACT 1946 struct mini_Qdisc __rcu *miniq_ingress; 1947 #endif 1948 struct netdev_queue __rcu *ingress_queue; 1949 #ifdef CONFIG_NETFILTER_INGRESS 1950 struct nf_hook_entries __rcu *nf_hooks_ingress; 1951 #endif 1952 1953 unsigned char broadcast[MAX_ADDR_LEN]; 1954 #ifdef CONFIG_RFS_ACCEL 1955 struct cpu_rmap *rx_cpu_rmap; 1956 #endif 1957 struct hlist_node index_hlist; 1958 1959 /* 1960 * Cache lines mostly used on transmit path 1961 */ 1962 struct netdev_queue *_tx ____cacheline_aligned_in_smp; 1963 unsigned int num_tx_queues; 1964 unsigned int real_num_tx_queues; 1965 struct Qdisc *qdisc; 1966 #ifdef CONFIG_NET_SCHED 1967 DECLARE_HASHTABLE (qdisc_hash, 4); 1968 #endif 1969 unsigned int tx_queue_len; 1970 spinlock_t tx_global_lock; 1971 int watchdog_timeo; 1972 1973 #ifdef CONFIG_XPS 1974 struct xps_dev_maps __rcu *xps_cpus_map; 1975 struct xps_dev_maps __rcu *xps_rxqs_map; 1976 #endif 1977 #ifdef CONFIG_NET_CLS_ACT 1978 struct mini_Qdisc __rcu *miniq_egress; 1979 #endif 1980 1981 /* These may be needed for future network-power-down code. */ 1982 struct timer_list watchdog_timer; 1983 1984 int __percpu *pcpu_refcnt; 1985 struct list_head todo_list; 1986 1987 struct list_head link_watch_list; 1988 1989 enum { NETREG_UNINITIALIZED=0, 1990 NETREG_REGISTERED, /* completed register_netdevice */ 1991 NETREG_UNREGISTERING, /* called unregister_netdevice */ 1992 NETREG_UNREGISTERED, /* completed unregister todo */ 1993 NETREG_RELEASED, /* called free_netdev */ 1994 NETREG_DUMMY, /* dummy device for NAPI poll */ 1995 } reg_state:8; 1996 1997 bool dismantle; 1998 1999 enum { 2000 RTNL_LINK_INITIALIZED, 2001 RTNL_LINK_INITIALIZING, 2002 } rtnl_link_state:16; 2003 2004 bool needs_free_netdev; 2005 void (*priv_destructor)(struct net_device *dev); 2006 2007 #ifdef CONFIG_NETPOLL 2008 struct netpoll_info __rcu *npinfo; 2009 #endif 2010 2011 possible_net_t nd_net; 2012 2013 /* mid-layer private */ 2014 union { 2015 void *ml_priv; 2016 struct pcpu_lstats __percpu *lstats; 2017 struct pcpu_sw_netstats __percpu *tstats; 2018 struct pcpu_dstats __percpu *dstats; 2019 }; 2020 2021 #if IS_ENABLED(CONFIG_GARP) 2022 struct garp_port __rcu *garp_port; 2023 #endif 2024 #if IS_ENABLED(CONFIG_MRP) 2025 struct mrp_port __rcu *mrp_port; 2026 #endif 2027 2028 struct device dev; 2029 const struct attribute_group *sysfs_groups[4]; 2030 const struct attribute_group *sysfs_rx_queue_group; 2031 2032 const struct rtnl_link_ops *rtnl_link_ops; 2033 2034 /* for setting kernel sock attribute on TCP connection setup */ 2035 #define GSO_MAX_SIZE 65536 2036 unsigned int gso_max_size; 2037 #define GSO_MAX_SEGS 65535 2038 u16 gso_max_segs; 2039 2040 #ifdef CONFIG_DCB 2041 const struct dcbnl_rtnl_ops *dcbnl_ops; 2042 #endif 2043 s16 num_tc; 2044 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE]; 2045 u8 prio_tc_map[TC_BITMASK + 1]; 2046 2047 #if IS_ENABLED(CONFIG_FCOE) 2048 unsigned int fcoe_ddp_xid; 2049 #endif 2050 #if IS_ENABLED(CONFIG_CGROUP_NET_PRIO) 2051 struct netprio_map __rcu *priomap; 2052 #endif 2053 struct phy_device *phydev; 2054 struct sfp_bus *sfp_bus; 2055 struct lock_class_key *qdisc_tx_busylock; 2056 struct lock_class_key *qdisc_running_key; 2057 bool proto_down; 2058 unsigned wol_enabled:1; 2059 }; 2060 #define to_net_dev(d) container_of(d, struct net_device, dev) 2061 2062 static inline bool netif_elide_gro(const struct net_device *dev) 2063 { 2064 if (!(dev->features & NETIF_F_GRO) || dev->xdp_prog) 2065 return true; 2066 return false; 2067 } 2068 2069 #define NETDEV_ALIGN 32 2070 2071 static inline 2072 int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio) 2073 { 2074 return dev->prio_tc_map[prio & TC_BITMASK]; 2075 } 2076 2077 static inline 2078 int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc) 2079 { 2080 if (tc >= dev->num_tc) 2081 return -EINVAL; 2082 2083 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK; 2084 return 0; 2085 } 2086 2087 int netdev_txq_to_tc(struct net_device *dev, unsigned int txq); 2088 void netdev_reset_tc(struct net_device *dev); 2089 int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset); 2090 int netdev_set_num_tc(struct net_device *dev, u8 num_tc); 2091 2092 static inline 2093 int netdev_get_num_tc(struct net_device *dev) 2094 { 2095 return dev->num_tc; 2096 } 2097 2098 void netdev_unbind_sb_channel(struct net_device *dev, 2099 struct net_device *sb_dev); 2100 int netdev_bind_sb_channel_queue(struct net_device *dev, 2101 struct net_device *sb_dev, 2102 u8 tc, u16 count, u16 offset); 2103 int netdev_set_sb_channel(struct net_device *dev, u16 channel); 2104 static inline int netdev_get_sb_channel(struct net_device *dev) 2105 { 2106 return max_t(int, -dev->num_tc, 0); 2107 } 2108 2109 static inline 2110 struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev, 2111 unsigned int index) 2112 { 2113 return &dev->_tx[index]; 2114 } 2115 2116 static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev, 2117 const struct sk_buff *skb) 2118 { 2119 return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb)); 2120 } 2121 2122 static inline void netdev_for_each_tx_queue(struct net_device *dev, 2123 void (*f)(struct net_device *, 2124 struct netdev_queue *, 2125 void *), 2126 void *arg) 2127 { 2128 unsigned int i; 2129 2130 for (i = 0; i < dev->num_tx_queues; i++) 2131 f(dev, &dev->_tx[i], arg); 2132 } 2133 2134 #define netdev_lockdep_set_classes(dev) \ 2135 { \ 2136 static struct lock_class_key qdisc_tx_busylock_key; \ 2137 static struct lock_class_key qdisc_running_key; \ 2138 static struct lock_class_key qdisc_xmit_lock_key; \ 2139 static struct lock_class_key dev_addr_list_lock_key; \ 2140 unsigned int i; \ 2141 \ 2142 (dev)->qdisc_tx_busylock = &qdisc_tx_busylock_key; \ 2143 (dev)->qdisc_running_key = &qdisc_running_key; \ 2144 lockdep_set_class(&(dev)->addr_list_lock, \ 2145 &dev_addr_list_lock_key); \ 2146 for (i = 0; i < (dev)->num_tx_queues; i++) \ 2147 lockdep_set_class(&(dev)->_tx[i]._xmit_lock, \ 2148 &qdisc_xmit_lock_key); \ 2149 } 2150 2151 struct netdev_queue *netdev_pick_tx(struct net_device *dev, 2152 struct sk_buff *skb, 2153 struct net_device *sb_dev); 2154 2155 /* returns the headroom that the master device needs to take in account 2156 * when forwarding to this dev 2157 */ 2158 static inline unsigned netdev_get_fwd_headroom(struct net_device *dev) 2159 { 2160 return dev->priv_flags & IFF_PHONY_HEADROOM ? 0 : dev->needed_headroom; 2161 } 2162 2163 static inline void netdev_set_rx_headroom(struct net_device *dev, int new_hr) 2164 { 2165 if (dev->netdev_ops->ndo_set_rx_headroom) 2166 dev->netdev_ops->ndo_set_rx_headroom(dev, new_hr); 2167 } 2168 2169 /* set the device rx headroom to the dev's default */ 2170 static inline void netdev_reset_rx_headroom(struct net_device *dev) 2171 { 2172 netdev_set_rx_headroom(dev, -1); 2173 } 2174 2175 /* 2176 * Net namespace inlines 2177 */ 2178 static inline 2179 struct net *dev_net(const struct net_device *dev) 2180 { 2181 return read_pnet(&dev->nd_net); 2182 } 2183 2184 static inline 2185 void dev_net_set(struct net_device *dev, struct net *net) 2186 { 2187 write_pnet(&dev->nd_net, net); 2188 } 2189 2190 /** 2191 * netdev_priv - access network device private data 2192 * @dev: network device 2193 * 2194 * Get network device private data 2195 */ 2196 static inline void *netdev_priv(const struct net_device *dev) 2197 { 2198 return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN); 2199 } 2200 2201 /* Set the sysfs physical device reference for the network logical device 2202 * if set prior to registration will cause a symlink during initialization. 2203 */ 2204 #define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev)) 2205 2206 /* Set the sysfs device type for the network logical device to allow 2207 * fine-grained identification of different network device types. For 2208 * example Ethernet, Wireless LAN, Bluetooth, WiMAX etc. 2209 */ 2210 #define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype)) 2211 2212 /* Default NAPI poll() weight 2213 * Device drivers are strongly advised to not use bigger value 2214 */ 2215 #define NAPI_POLL_WEIGHT 64 2216 2217 /** 2218 * netif_napi_add - initialize a NAPI context 2219 * @dev: network device 2220 * @napi: NAPI context 2221 * @poll: polling function 2222 * @weight: default weight 2223 * 2224 * netif_napi_add() must be used to initialize a NAPI context prior to calling 2225 * *any* of the other NAPI-related functions. 2226 */ 2227 void netif_napi_add(struct net_device *dev, struct napi_struct *napi, 2228 int (*poll)(struct napi_struct *, int), int weight); 2229 2230 /** 2231 * netif_tx_napi_add - initialize a NAPI context 2232 * @dev: network device 2233 * @napi: NAPI context 2234 * @poll: polling function 2235 * @weight: default weight 2236 * 2237 * This variant of netif_napi_add() should be used from drivers using NAPI 2238 * to exclusively poll a TX queue. 2239 * This will avoid we add it into napi_hash[], thus polluting this hash table. 2240 */ 2241 static inline void netif_tx_napi_add(struct net_device *dev, 2242 struct napi_struct *napi, 2243 int (*poll)(struct napi_struct *, int), 2244 int weight) 2245 { 2246 set_bit(NAPI_STATE_NO_BUSY_POLL, &napi->state); 2247 netif_napi_add(dev, napi, poll, weight); 2248 } 2249 2250 /** 2251 * netif_napi_del - remove a NAPI context 2252 * @napi: NAPI context 2253 * 2254 * netif_napi_del() removes a NAPI context from the network device NAPI list 2255 */ 2256 void netif_napi_del(struct napi_struct *napi); 2257 2258 struct napi_gro_cb { 2259 /* Virtual address of skb_shinfo(skb)->frags[0].page + offset. */ 2260 void *frag0; 2261 2262 /* Length of frag0. */ 2263 unsigned int frag0_len; 2264 2265 /* This indicates where we are processing relative to skb->data. */ 2266 int data_offset; 2267 2268 /* This is non-zero if the packet cannot be merged with the new skb. */ 2269 u16 flush; 2270 2271 /* Save the IP ID here and check when we get to the transport layer */ 2272 u16 flush_id; 2273 2274 /* Number of segments aggregated. */ 2275 u16 count; 2276 2277 /* Start offset for remote checksum offload */ 2278 u16 gro_remcsum_start; 2279 2280 /* jiffies when first packet was created/queued */ 2281 unsigned long age; 2282 2283 /* Used in ipv6_gro_receive() and foo-over-udp */ 2284 u16 proto; 2285 2286 /* This is non-zero if the packet may be of the same flow. */ 2287 u8 same_flow:1; 2288 2289 /* Used in tunnel GRO receive */ 2290 u8 encap_mark:1; 2291 2292 /* GRO checksum is valid */ 2293 u8 csum_valid:1; 2294 2295 /* Number of checksums via CHECKSUM_UNNECESSARY */ 2296 u8 csum_cnt:3; 2297 2298 /* Free the skb? */ 2299 u8 free:2; 2300 #define NAPI_GRO_FREE 1 2301 #define NAPI_GRO_FREE_STOLEN_HEAD 2 2302 2303 /* Used in foo-over-udp, set in udp[46]_gro_receive */ 2304 u8 is_ipv6:1; 2305 2306 /* Used in GRE, set in fou/gue_gro_receive */ 2307 u8 is_fou:1; 2308 2309 /* Used to determine if flush_id can be ignored */ 2310 u8 is_atomic:1; 2311 2312 /* Number of gro_receive callbacks this packet already went through */ 2313 u8 recursion_counter:4; 2314 2315 /* 1 bit hole */ 2316 2317 /* used to support CHECKSUM_COMPLETE for tunneling protocols */ 2318 __wsum csum; 2319 2320 /* used in skb_gro_receive() slow path */ 2321 struct sk_buff *last; 2322 }; 2323 2324 #define NAPI_GRO_CB(skb) ((struct napi_gro_cb *)(skb)->cb) 2325 2326 #define GRO_RECURSION_LIMIT 15 2327 static inline int gro_recursion_inc_test(struct sk_buff *skb) 2328 { 2329 return ++NAPI_GRO_CB(skb)->recursion_counter == GRO_RECURSION_LIMIT; 2330 } 2331 2332 typedef struct sk_buff *(*gro_receive_t)(struct list_head *, struct sk_buff *); 2333 static inline struct sk_buff *call_gro_receive(gro_receive_t cb, 2334 struct list_head *head, 2335 struct sk_buff *skb) 2336 { 2337 if (unlikely(gro_recursion_inc_test(skb))) { 2338 NAPI_GRO_CB(skb)->flush |= 1; 2339 return NULL; 2340 } 2341 2342 return cb(head, skb); 2343 } 2344 2345 typedef struct sk_buff *(*gro_receive_sk_t)(struct sock *, struct list_head *, 2346 struct sk_buff *); 2347 static inline struct sk_buff *call_gro_receive_sk(gro_receive_sk_t cb, 2348 struct sock *sk, 2349 struct list_head *head, 2350 struct sk_buff *skb) 2351 { 2352 if (unlikely(gro_recursion_inc_test(skb))) { 2353 NAPI_GRO_CB(skb)->flush |= 1; 2354 return NULL; 2355 } 2356 2357 return cb(sk, head, skb); 2358 } 2359 2360 struct packet_type { 2361 __be16 type; /* This is really htons(ether_type). */ 2362 bool ignore_outgoing; 2363 struct net_device *dev; /* NULL is wildcarded here */ 2364 int (*func) (struct sk_buff *, 2365 struct net_device *, 2366 struct packet_type *, 2367 struct net_device *); 2368 void (*list_func) (struct list_head *, 2369 struct packet_type *, 2370 struct net_device *); 2371 bool (*id_match)(struct packet_type *ptype, 2372 struct sock *sk); 2373 void *af_packet_priv; 2374 struct list_head list; 2375 }; 2376 2377 struct offload_callbacks { 2378 struct sk_buff *(*gso_segment)(struct sk_buff *skb, 2379 netdev_features_t features); 2380 struct sk_buff *(*gro_receive)(struct list_head *head, 2381 struct sk_buff *skb); 2382 int (*gro_complete)(struct sk_buff *skb, int nhoff); 2383 }; 2384 2385 struct packet_offload { 2386 __be16 type; /* This is really htons(ether_type). */ 2387 u16 priority; 2388 struct offload_callbacks callbacks; 2389 struct list_head list; 2390 }; 2391 2392 /* often modified stats are per-CPU, other are shared (netdev->stats) */ 2393 struct pcpu_sw_netstats { 2394 u64 rx_packets; 2395 u64 rx_bytes; 2396 u64 tx_packets; 2397 u64 tx_bytes; 2398 struct u64_stats_sync syncp; 2399 } __aligned(4 * sizeof(u64)); 2400 2401 struct pcpu_lstats { 2402 u64 packets; 2403 u64 bytes; 2404 struct u64_stats_sync syncp; 2405 } __aligned(2 * sizeof(u64)); 2406 2407 #define __netdev_alloc_pcpu_stats(type, gfp) \ 2408 ({ \ 2409 typeof(type) __percpu *pcpu_stats = alloc_percpu_gfp(type, gfp);\ 2410 if (pcpu_stats) { \ 2411 int __cpu; \ 2412 for_each_possible_cpu(__cpu) { \ 2413 typeof(type) *stat; \ 2414 stat = per_cpu_ptr(pcpu_stats, __cpu); \ 2415 u64_stats_init(&stat->syncp); \ 2416 } \ 2417 } \ 2418 pcpu_stats; \ 2419 }) 2420 2421 #define netdev_alloc_pcpu_stats(type) \ 2422 __netdev_alloc_pcpu_stats(type, GFP_KERNEL) 2423 2424 enum netdev_lag_tx_type { 2425 NETDEV_LAG_TX_TYPE_UNKNOWN, 2426 NETDEV_LAG_TX_TYPE_RANDOM, 2427 NETDEV_LAG_TX_TYPE_BROADCAST, 2428 NETDEV_LAG_TX_TYPE_ROUNDROBIN, 2429 NETDEV_LAG_TX_TYPE_ACTIVEBACKUP, 2430 NETDEV_LAG_TX_TYPE_HASH, 2431 }; 2432 2433 enum netdev_lag_hash { 2434 NETDEV_LAG_HASH_NONE, 2435 NETDEV_LAG_HASH_L2, 2436 NETDEV_LAG_HASH_L34, 2437 NETDEV_LAG_HASH_L23, 2438 NETDEV_LAG_HASH_E23, 2439 NETDEV_LAG_HASH_E34, 2440 NETDEV_LAG_HASH_UNKNOWN, 2441 }; 2442 2443 struct netdev_lag_upper_info { 2444 enum netdev_lag_tx_type tx_type; 2445 enum netdev_lag_hash hash_type; 2446 }; 2447 2448 struct netdev_lag_lower_state_info { 2449 u8 link_up : 1, 2450 tx_enabled : 1; 2451 }; 2452 2453 #include <linux/notifier.h> 2454 2455 /* netdevice notifier chain. Please remember to update netdev_cmd_to_name() 2456 * and the rtnetlink notification exclusion list in rtnetlink_event() when 2457 * adding new types. 2458 */ 2459 enum netdev_cmd { 2460 NETDEV_UP = 1, /* For now you can't veto a device up/down */ 2461 NETDEV_DOWN, 2462 NETDEV_REBOOT, /* Tell a protocol stack a network interface 2463 detected a hardware crash and restarted 2464 - we can use this eg to kick tcp sessions 2465 once done */ 2466 NETDEV_CHANGE, /* Notify device state change */ 2467 NETDEV_REGISTER, 2468 NETDEV_UNREGISTER, 2469 NETDEV_CHANGEMTU, /* notify after mtu change happened */ 2470 NETDEV_CHANGEADDR, /* notify after the address change */ 2471 NETDEV_PRE_CHANGEADDR, /* notify before the address change */ 2472 NETDEV_GOING_DOWN, 2473 NETDEV_CHANGENAME, 2474 NETDEV_FEAT_CHANGE, 2475 NETDEV_BONDING_FAILOVER, 2476 NETDEV_PRE_UP, 2477 NETDEV_PRE_TYPE_CHANGE, 2478 NETDEV_POST_TYPE_CHANGE, 2479 NETDEV_POST_INIT, 2480 NETDEV_RELEASE, 2481 NETDEV_NOTIFY_PEERS, 2482 NETDEV_JOIN, 2483 NETDEV_CHANGEUPPER, 2484 NETDEV_RESEND_IGMP, 2485 NETDEV_PRECHANGEMTU, /* notify before mtu change happened */ 2486 NETDEV_CHANGEINFODATA, 2487 NETDEV_BONDING_INFO, 2488 NETDEV_PRECHANGEUPPER, 2489 NETDEV_CHANGELOWERSTATE, 2490 NETDEV_UDP_TUNNEL_PUSH_INFO, 2491 NETDEV_UDP_TUNNEL_DROP_INFO, 2492 NETDEV_CHANGE_TX_QUEUE_LEN, 2493 NETDEV_CVLAN_FILTER_PUSH_INFO, 2494 NETDEV_CVLAN_FILTER_DROP_INFO, 2495 NETDEV_SVLAN_FILTER_PUSH_INFO, 2496 NETDEV_SVLAN_FILTER_DROP_INFO, 2497 }; 2498 const char *netdev_cmd_to_name(enum netdev_cmd cmd); 2499 2500 int register_netdevice_notifier(struct notifier_block *nb); 2501 int unregister_netdevice_notifier(struct notifier_block *nb); 2502 2503 struct netdev_notifier_info { 2504 struct net_device *dev; 2505 struct netlink_ext_ack *extack; 2506 }; 2507 2508 struct netdev_notifier_info_ext { 2509 struct netdev_notifier_info info; /* must be first */ 2510 union { 2511 u32 mtu; 2512 } ext; 2513 }; 2514 2515 struct netdev_notifier_change_info { 2516 struct netdev_notifier_info info; /* must be first */ 2517 unsigned int flags_changed; 2518 }; 2519 2520 struct netdev_notifier_changeupper_info { 2521 struct netdev_notifier_info info; /* must be first */ 2522 struct net_device *upper_dev; /* new upper dev */ 2523 bool master; /* is upper dev master */ 2524 bool linking; /* is the notification for link or unlink */ 2525 void *upper_info; /* upper dev info */ 2526 }; 2527 2528 struct netdev_notifier_changelowerstate_info { 2529 struct netdev_notifier_info info; /* must be first */ 2530 void *lower_state_info; /* is lower dev state */ 2531 }; 2532 2533 struct netdev_notifier_pre_changeaddr_info { 2534 struct netdev_notifier_info info; /* must be first */ 2535 const unsigned char *dev_addr; 2536 }; 2537 2538 static inline void netdev_notifier_info_init(struct netdev_notifier_info *info, 2539 struct net_device *dev) 2540 { 2541 info->dev = dev; 2542 info->extack = NULL; 2543 } 2544 2545 static inline struct net_device * 2546 netdev_notifier_info_to_dev(const struct netdev_notifier_info *info) 2547 { 2548 return info->dev; 2549 } 2550 2551 static inline struct netlink_ext_ack * 2552 netdev_notifier_info_to_extack(const struct netdev_notifier_info *info) 2553 { 2554 return info->extack; 2555 } 2556 2557 int call_netdevice_notifiers(unsigned long val, struct net_device *dev); 2558 2559 2560 extern rwlock_t dev_base_lock; /* Device list lock */ 2561 2562 #define for_each_netdev(net, d) \ 2563 list_for_each_entry(d, &(net)->dev_base_head, dev_list) 2564 #define for_each_netdev_reverse(net, d) \ 2565 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list) 2566 #define for_each_netdev_rcu(net, d) \ 2567 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list) 2568 #define for_each_netdev_safe(net, d, n) \ 2569 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list) 2570 #define for_each_netdev_continue(net, d) \ 2571 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list) 2572 #define for_each_netdev_continue_rcu(net, d) \ 2573 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list) 2574 #define for_each_netdev_in_bond_rcu(bond, slave) \ 2575 for_each_netdev_rcu(&init_net, slave) \ 2576 if (netdev_master_upper_dev_get_rcu(slave) == (bond)) 2577 #define net_device_entry(lh) list_entry(lh, struct net_device, dev_list) 2578 2579 static inline struct net_device *next_net_device(struct net_device *dev) 2580 { 2581 struct list_head *lh; 2582 struct net *net; 2583 2584 net = dev_net(dev); 2585 lh = dev->dev_list.next; 2586 return lh == &net->dev_base_head ? NULL : net_device_entry(lh); 2587 } 2588 2589 static inline struct net_device *next_net_device_rcu(struct net_device *dev) 2590 { 2591 struct list_head *lh; 2592 struct net *net; 2593 2594 net = dev_net(dev); 2595 lh = rcu_dereference(list_next_rcu(&dev->dev_list)); 2596 return lh == &net->dev_base_head ? NULL : net_device_entry(lh); 2597 } 2598 2599 static inline struct net_device *first_net_device(struct net *net) 2600 { 2601 return list_empty(&net->dev_base_head) ? NULL : 2602 net_device_entry(net->dev_base_head.next); 2603 } 2604 2605 static inline struct net_device *first_net_device_rcu(struct net *net) 2606 { 2607 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head)); 2608 2609 return lh == &net->dev_base_head ? NULL : net_device_entry(lh); 2610 } 2611 2612 int netdev_boot_setup_check(struct net_device *dev); 2613 unsigned long netdev_boot_base(const char *prefix, int unit); 2614 struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type, 2615 const char *hwaddr); 2616 struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type); 2617 struct net_device *__dev_getfirstbyhwtype(struct net *net, unsigned short type); 2618 void dev_add_pack(struct packet_type *pt); 2619 void dev_remove_pack(struct packet_type *pt); 2620 void __dev_remove_pack(struct packet_type *pt); 2621 void dev_add_offload(struct packet_offload *po); 2622 void dev_remove_offload(struct packet_offload *po); 2623 2624 int dev_get_iflink(const struct net_device *dev); 2625 int dev_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb); 2626 struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags, 2627 unsigned short mask); 2628 struct net_device *dev_get_by_name(struct net *net, const char *name); 2629 struct net_device *dev_get_by_name_rcu(struct net *net, const char *name); 2630 struct net_device *__dev_get_by_name(struct net *net, const char *name); 2631 int dev_alloc_name(struct net_device *dev, const char *name); 2632 int dev_open(struct net_device *dev, struct netlink_ext_ack *extack); 2633 void dev_close(struct net_device *dev); 2634 void dev_close_many(struct list_head *head, bool unlink); 2635 void dev_disable_lro(struct net_device *dev); 2636 int dev_loopback_xmit(struct net *net, struct sock *sk, struct sk_buff *newskb); 2637 u16 dev_pick_tx_zero(struct net_device *dev, struct sk_buff *skb, 2638 struct net_device *sb_dev, 2639 select_queue_fallback_t fallback); 2640 u16 dev_pick_tx_cpu_id(struct net_device *dev, struct sk_buff *skb, 2641 struct net_device *sb_dev, 2642 select_queue_fallback_t fallback); 2643 int dev_queue_xmit(struct sk_buff *skb); 2644 int dev_queue_xmit_accel(struct sk_buff *skb, struct net_device *sb_dev); 2645 int dev_direct_xmit(struct sk_buff *skb, u16 queue_id); 2646 int register_netdevice(struct net_device *dev); 2647 void unregister_netdevice_queue(struct net_device *dev, struct list_head *head); 2648 void unregister_netdevice_many(struct list_head *head); 2649 static inline void unregister_netdevice(struct net_device *dev) 2650 { 2651 unregister_netdevice_queue(dev, NULL); 2652 } 2653 2654 int netdev_refcnt_read(const struct net_device *dev); 2655 void free_netdev(struct net_device *dev); 2656 void netdev_freemem(struct net_device *dev); 2657 void synchronize_net(void); 2658 int init_dummy_netdev(struct net_device *dev); 2659 2660 DECLARE_PER_CPU(int, xmit_recursion); 2661 #define XMIT_RECURSION_LIMIT 10 2662 2663 static inline int dev_recursion_level(void) 2664 { 2665 return this_cpu_read(xmit_recursion); 2666 } 2667 2668 struct net_device *dev_get_by_index(struct net *net, int ifindex); 2669 struct net_device *__dev_get_by_index(struct net *net, int ifindex); 2670 struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex); 2671 struct net_device *dev_get_by_napi_id(unsigned int napi_id); 2672 int netdev_get_name(struct net *net, char *name, int ifindex); 2673 int dev_restart(struct net_device *dev); 2674 int skb_gro_receive(struct sk_buff *p, struct sk_buff *skb); 2675 2676 static inline unsigned int skb_gro_offset(const struct sk_buff *skb) 2677 { 2678 return NAPI_GRO_CB(skb)->data_offset; 2679 } 2680 2681 static inline unsigned int skb_gro_len(const struct sk_buff *skb) 2682 { 2683 return skb->len - NAPI_GRO_CB(skb)->data_offset; 2684 } 2685 2686 static inline void skb_gro_pull(struct sk_buff *skb, unsigned int len) 2687 { 2688 NAPI_GRO_CB(skb)->data_offset += len; 2689 } 2690 2691 static inline void *skb_gro_header_fast(struct sk_buff *skb, 2692 unsigned int offset) 2693 { 2694 return NAPI_GRO_CB(skb)->frag0 + offset; 2695 } 2696 2697 static inline int skb_gro_header_hard(struct sk_buff *skb, unsigned int hlen) 2698 { 2699 return NAPI_GRO_CB(skb)->frag0_len < hlen; 2700 } 2701 2702 static inline void skb_gro_frag0_invalidate(struct sk_buff *skb) 2703 { 2704 NAPI_GRO_CB(skb)->frag0 = NULL; 2705 NAPI_GRO_CB(skb)->frag0_len = 0; 2706 } 2707 2708 static inline void *skb_gro_header_slow(struct sk_buff *skb, unsigned int hlen, 2709 unsigned int offset) 2710 { 2711 if (!pskb_may_pull(skb, hlen)) 2712 return NULL; 2713 2714 skb_gro_frag0_invalidate(skb); 2715 return skb->data + offset; 2716 } 2717 2718 static inline void *skb_gro_network_header(struct sk_buff *skb) 2719 { 2720 return (NAPI_GRO_CB(skb)->frag0 ?: skb->data) + 2721 skb_network_offset(skb); 2722 } 2723 2724 static inline void skb_gro_postpull_rcsum(struct sk_buff *skb, 2725 const void *start, unsigned int len) 2726 { 2727 if (NAPI_GRO_CB(skb)->csum_valid) 2728 NAPI_GRO_CB(skb)->csum = csum_sub(NAPI_GRO_CB(skb)->csum, 2729 csum_partial(start, len, 0)); 2730 } 2731 2732 /* GRO checksum functions. These are logical equivalents of the normal 2733 * checksum functions (in skbuff.h) except that they operate on the GRO 2734 * offsets and fields in sk_buff. 2735 */ 2736 2737 __sum16 __skb_gro_checksum_complete(struct sk_buff *skb); 2738 2739 static inline bool skb_at_gro_remcsum_start(struct sk_buff *skb) 2740 { 2741 return (NAPI_GRO_CB(skb)->gro_remcsum_start == skb_gro_offset(skb)); 2742 } 2743 2744 static inline bool __skb_gro_checksum_validate_needed(struct sk_buff *skb, 2745 bool zero_okay, 2746 __sum16 check) 2747 { 2748 return ((skb->ip_summed != CHECKSUM_PARTIAL || 2749 skb_checksum_start_offset(skb) < 2750 skb_gro_offset(skb)) && 2751 !skb_at_gro_remcsum_start(skb) && 2752 NAPI_GRO_CB(skb)->csum_cnt == 0 && 2753 (!zero_okay || check)); 2754 } 2755 2756 static inline __sum16 __skb_gro_checksum_validate_complete(struct sk_buff *skb, 2757 __wsum psum) 2758 { 2759 if (NAPI_GRO_CB(skb)->csum_valid && 2760 !csum_fold(csum_add(psum, NAPI_GRO_CB(skb)->csum))) 2761 return 0; 2762 2763 NAPI_GRO_CB(skb)->csum = psum; 2764 2765 return __skb_gro_checksum_complete(skb); 2766 } 2767 2768 static inline void skb_gro_incr_csum_unnecessary(struct sk_buff *skb) 2769 { 2770 if (NAPI_GRO_CB(skb)->csum_cnt > 0) { 2771 /* Consume a checksum from CHECKSUM_UNNECESSARY */ 2772 NAPI_GRO_CB(skb)->csum_cnt--; 2773 } else { 2774 /* Update skb for CHECKSUM_UNNECESSARY and csum_level when we 2775 * verified a new top level checksum or an encapsulated one 2776 * during GRO. This saves work if we fallback to normal path. 2777 */ 2778 __skb_incr_checksum_unnecessary(skb); 2779 } 2780 } 2781 2782 #define __skb_gro_checksum_validate(skb, proto, zero_okay, check, \ 2783 compute_pseudo) \ 2784 ({ \ 2785 __sum16 __ret = 0; \ 2786 if (__skb_gro_checksum_validate_needed(skb, zero_okay, check)) \ 2787 __ret = __skb_gro_checksum_validate_complete(skb, \ 2788 compute_pseudo(skb, proto)); \ 2789 if (!__ret) \ 2790 skb_gro_incr_csum_unnecessary(skb); \ 2791 __ret; \ 2792 }) 2793 2794 #define skb_gro_checksum_validate(skb, proto, compute_pseudo) \ 2795 __skb_gro_checksum_validate(skb, proto, false, 0, compute_pseudo) 2796 2797 #define skb_gro_checksum_validate_zero_check(skb, proto, check, \ 2798 compute_pseudo) \ 2799 __skb_gro_checksum_validate(skb, proto, true, check, compute_pseudo) 2800 2801 #define skb_gro_checksum_simple_validate(skb) \ 2802 __skb_gro_checksum_validate(skb, 0, false, 0, null_compute_pseudo) 2803 2804 static inline bool __skb_gro_checksum_convert_check(struct sk_buff *skb) 2805 { 2806 return (NAPI_GRO_CB(skb)->csum_cnt == 0 && 2807 !NAPI_GRO_CB(skb)->csum_valid); 2808 } 2809 2810 static inline void __skb_gro_checksum_convert(struct sk_buff *skb, 2811 __sum16 check, __wsum pseudo) 2812 { 2813 NAPI_GRO_CB(skb)->csum = ~pseudo; 2814 NAPI_GRO_CB(skb)->csum_valid = 1; 2815 } 2816 2817 #define skb_gro_checksum_try_convert(skb, proto, check, compute_pseudo) \ 2818 do { \ 2819 if (__skb_gro_checksum_convert_check(skb)) \ 2820 __skb_gro_checksum_convert(skb, check, \ 2821 compute_pseudo(skb, proto)); \ 2822 } while (0) 2823 2824 struct gro_remcsum { 2825 int offset; 2826 __wsum delta; 2827 }; 2828 2829 static inline void skb_gro_remcsum_init(struct gro_remcsum *grc) 2830 { 2831 grc->offset = 0; 2832 grc->delta = 0; 2833 } 2834 2835 static inline void *skb_gro_remcsum_process(struct sk_buff *skb, void *ptr, 2836 unsigned int off, size_t hdrlen, 2837 int start, int offset, 2838 struct gro_remcsum *grc, 2839 bool nopartial) 2840 { 2841 __wsum delta; 2842 size_t plen = hdrlen + max_t(size_t, offset + sizeof(u16), start); 2843 2844 BUG_ON(!NAPI_GRO_CB(skb)->csum_valid); 2845 2846 if (!nopartial) { 2847 NAPI_GRO_CB(skb)->gro_remcsum_start = off + hdrlen + start; 2848 return ptr; 2849 } 2850 2851 ptr = skb_gro_header_fast(skb, off); 2852 if (skb_gro_header_hard(skb, off + plen)) { 2853 ptr = skb_gro_header_slow(skb, off + plen, off); 2854 if (!ptr) 2855 return NULL; 2856 } 2857 2858 delta = remcsum_adjust(ptr + hdrlen, NAPI_GRO_CB(skb)->csum, 2859 start, offset); 2860 2861 /* Adjust skb->csum since we changed the packet */ 2862 NAPI_GRO_CB(skb)->csum = csum_add(NAPI_GRO_CB(skb)->csum, delta); 2863 2864 grc->offset = off + hdrlen + offset; 2865 grc->delta = delta; 2866 2867 return ptr; 2868 } 2869 2870 static inline void skb_gro_remcsum_cleanup(struct sk_buff *skb, 2871 struct gro_remcsum *grc) 2872 { 2873 void *ptr; 2874 size_t plen = grc->offset + sizeof(u16); 2875 2876 if (!grc->delta) 2877 return; 2878 2879 ptr = skb_gro_header_fast(skb, grc->offset); 2880 if (skb_gro_header_hard(skb, grc->offset + sizeof(u16))) { 2881 ptr = skb_gro_header_slow(skb, plen, grc->offset); 2882 if (!ptr) 2883 return; 2884 } 2885 2886 remcsum_unadjust((__sum16 *)ptr, grc->delta); 2887 } 2888 2889 #ifdef CONFIG_XFRM_OFFLOAD 2890 static inline void skb_gro_flush_final(struct sk_buff *skb, struct sk_buff *pp, int flush) 2891 { 2892 if (PTR_ERR(pp) != -EINPROGRESS) 2893 NAPI_GRO_CB(skb)->flush |= flush; 2894 } 2895 static inline void skb_gro_flush_final_remcsum(struct sk_buff *skb, 2896 struct sk_buff *pp, 2897 int flush, 2898 struct gro_remcsum *grc) 2899 { 2900 if (PTR_ERR(pp) != -EINPROGRESS) { 2901 NAPI_GRO_CB(skb)->flush |= flush; 2902 skb_gro_remcsum_cleanup(skb, grc); 2903 skb->remcsum_offload = 0; 2904 } 2905 } 2906 #else 2907 static inline void skb_gro_flush_final(struct sk_buff *skb, struct sk_buff *pp, int flush) 2908 { 2909 NAPI_GRO_CB(skb)->flush |= flush; 2910 } 2911 static inline void skb_gro_flush_final_remcsum(struct sk_buff *skb, 2912 struct sk_buff *pp, 2913 int flush, 2914 struct gro_remcsum *grc) 2915 { 2916 NAPI_GRO_CB(skb)->flush |= flush; 2917 skb_gro_remcsum_cleanup(skb, grc); 2918 skb->remcsum_offload = 0; 2919 } 2920 #endif 2921 2922 static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev, 2923 unsigned short type, 2924 const void *daddr, const void *saddr, 2925 unsigned int len) 2926 { 2927 if (!dev->header_ops || !dev->header_ops->create) 2928 return 0; 2929 2930 return dev->header_ops->create(skb, dev, type, daddr, saddr, len); 2931 } 2932 2933 static inline int dev_parse_header(const struct sk_buff *skb, 2934 unsigned char *haddr) 2935 { 2936 const struct net_device *dev = skb->dev; 2937 2938 if (!dev->header_ops || !dev->header_ops->parse) 2939 return 0; 2940 return dev->header_ops->parse(skb, haddr); 2941 } 2942 2943 /* ll_header must have at least hard_header_len allocated */ 2944 static inline bool dev_validate_header(const struct net_device *dev, 2945 char *ll_header, int len) 2946 { 2947 if (likely(len >= dev->hard_header_len)) 2948 return true; 2949 if (len < dev->min_header_len) 2950 return false; 2951 2952 if (capable(CAP_SYS_RAWIO)) { 2953 memset(ll_header + len, 0, dev->hard_header_len - len); 2954 return true; 2955 } 2956 2957 if (dev->header_ops && dev->header_ops->validate) 2958 return dev->header_ops->validate(ll_header, len); 2959 2960 return false; 2961 } 2962 2963 typedef int gifconf_func_t(struct net_device * dev, char __user * bufptr, 2964 int len, int size); 2965 int register_gifconf(unsigned int family, gifconf_func_t *gifconf); 2966 static inline int unregister_gifconf(unsigned int family) 2967 { 2968 return register_gifconf(family, NULL); 2969 } 2970 2971 #ifdef CONFIG_NET_FLOW_LIMIT 2972 #define FLOW_LIMIT_HISTORY (1 << 7) /* must be ^2 and !overflow buckets */ 2973 struct sd_flow_limit { 2974 u64 count; 2975 unsigned int num_buckets; 2976 unsigned int history_head; 2977 u16 history[FLOW_LIMIT_HISTORY]; 2978 u8 buckets[]; 2979 }; 2980 2981 extern int netdev_flow_limit_table_len; 2982 #endif /* CONFIG_NET_FLOW_LIMIT */ 2983 2984 /* 2985 * Incoming packets are placed on per-CPU queues 2986 */ 2987 struct softnet_data { 2988 struct list_head poll_list; 2989 struct sk_buff_head process_queue; 2990 2991 /* stats */ 2992 unsigned int processed; 2993 unsigned int time_squeeze; 2994 unsigned int received_rps; 2995 #ifdef CONFIG_RPS 2996 struct softnet_data *rps_ipi_list; 2997 #endif 2998 #ifdef CONFIG_NET_FLOW_LIMIT 2999 struct sd_flow_limit __rcu *flow_limit; 3000 #endif 3001 struct Qdisc *output_queue; 3002 struct Qdisc **output_queue_tailp; 3003 struct sk_buff *completion_queue; 3004 #ifdef CONFIG_XFRM_OFFLOAD 3005 struct sk_buff_head xfrm_backlog; 3006 #endif 3007 #ifdef CONFIG_RPS 3008 /* input_queue_head should be written by cpu owning this struct, 3009 * and only read by other cpus. Worth using a cache line. 3010 */ 3011 unsigned int input_queue_head ____cacheline_aligned_in_smp; 3012 3013 /* Elements below can be accessed between CPUs for RPS/RFS */ 3014 call_single_data_t csd ____cacheline_aligned_in_smp; 3015 struct softnet_data *rps_ipi_next; 3016 unsigned int cpu; 3017 unsigned int input_queue_tail; 3018 #endif 3019 unsigned int dropped; 3020 struct sk_buff_head input_pkt_queue; 3021 struct napi_struct backlog; 3022 3023 }; 3024 3025 static inline void input_queue_head_incr(struct softnet_data *sd) 3026 { 3027 #ifdef CONFIG_RPS 3028 sd->input_queue_head++; 3029 #endif 3030 } 3031 3032 static inline void input_queue_tail_incr_save(struct softnet_data *sd, 3033 unsigned int *qtail) 3034 { 3035 #ifdef CONFIG_RPS 3036 *qtail = ++sd->input_queue_tail; 3037 #endif 3038 } 3039 3040 DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data); 3041 3042 void __netif_schedule(struct Qdisc *q); 3043 void netif_schedule_queue(struct netdev_queue *txq); 3044 3045 static inline void netif_tx_schedule_all(struct net_device *dev) 3046 { 3047 unsigned int i; 3048 3049 for (i = 0; i < dev->num_tx_queues; i++) 3050 netif_schedule_queue(netdev_get_tx_queue(dev, i)); 3051 } 3052 3053 static __always_inline void netif_tx_start_queue(struct netdev_queue *dev_queue) 3054 { 3055 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state); 3056 } 3057 3058 /** 3059 * netif_start_queue - allow transmit 3060 * @dev: network device 3061 * 3062 * Allow upper layers to call the device hard_start_xmit routine. 3063 */ 3064 static inline void netif_start_queue(struct net_device *dev) 3065 { 3066 netif_tx_start_queue(netdev_get_tx_queue(dev, 0)); 3067 } 3068 3069 static inline void netif_tx_start_all_queues(struct net_device *dev) 3070 { 3071 unsigned int i; 3072 3073 for (i = 0; i < dev->num_tx_queues; i++) { 3074 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 3075 netif_tx_start_queue(txq); 3076 } 3077 } 3078 3079 void netif_tx_wake_queue(struct netdev_queue *dev_queue); 3080 3081 /** 3082 * netif_wake_queue - restart transmit 3083 * @dev: network device 3084 * 3085 * Allow upper layers to call the device hard_start_xmit routine. 3086 * Used for flow control when transmit resources are available. 3087 */ 3088 static inline void netif_wake_queue(struct net_device *dev) 3089 { 3090 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0)); 3091 } 3092 3093 static inline void netif_tx_wake_all_queues(struct net_device *dev) 3094 { 3095 unsigned int i; 3096 3097 for (i = 0; i < dev->num_tx_queues; i++) { 3098 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 3099 netif_tx_wake_queue(txq); 3100 } 3101 } 3102 3103 static __always_inline void netif_tx_stop_queue(struct netdev_queue *dev_queue) 3104 { 3105 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state); 3106 } 3107 3108 /** 3109 * netif_stop_queue - stop transmitted packets 3110 * @dev: network device 3111 * 3112 * Stop upper layers calling the device hard_start_xmit routine. 3113 * Used for flow control when transmit resources are unavailable. 3114 */ 3115 static inline void netif_stop_queue(struct net_device *dev) 3116 { 3117 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0)); 3118 } 3119 3120 void netif_tx_stop_all_queues(struct net_device *dev); 3121 3122 static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue) 3123 { 3124 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state); 3125 } 3126 3127 /** 3128 * netif_queue_stopped - test if transmit queue is flowblocked 3129 * @dev: network device 3130 * 3131 * Test if transmit queue on device is currently unable to send. 3132 */ 3133 static inline bool netif_queue_stopped(const struct net_device *dev) 3134 { 3135 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0)); 3136 } 3137 3138 static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue) 3139 { 3140 return dev_queue->state & QUEUE_STATE_ANY_XOFF; 3141 } 3142 3143 static inline bool 3144 netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue) 3145 { 3146 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN; 3147 } 3148 3149 static inline bool 3150 netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue) 3151 { 3152 return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN; 3153 } 3154 3155 /** 3156 * netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write 3157 * @dev_queue: pointer to transmit queue 3158 * 3159 * BQL enabled drivers might use this helper in their ndo_start_xmit(), 3160 * to give appropriate hint to the CPU. 3161 */ 3162 static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue) 3163 { 3164 #ifdef CONFIG_BQL 3165 prefetchw(&dev_queue->dql.num_queued); 3166 #endif 3167 } 3168 3169 /** 3170 * netdev_txq_bql_complete_prefetchw - prefetch bql data for write 3171 * @dev_queue: pointer to transmit queue 3172 * 3173 * BQL enabled drivers might use this helper in their TX completion path, 3174 * to give appropriate hint to the CPU. 3175 */ 3176 static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue) 3177 { 3178 #ifdef CONFIG_BQL 3179 prefetchw(&dev_queue->dql.limit); 3180 #endif 3181 } 3182 3183 static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue, 3184 unsigned int bytes) 3185 { 3186 #ifdef CONFIG_BQL 3187 dql_queued(&dev_queue->dql, bytes); 3188 3189 if (likely(dql_avail(&dev_queue->dql) >= 0)) 3190 return; 3191 3192 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state); 3193 3194 /* 3195 * The XOFF flag must be set before checking the dql_avail below, 3196 * because in netdev_tx_completed_queue we update the dql_completed 3197 * before checking the XOFF flag. 3198 */ 3199 smp_mb(); 3200 3201 /* check again in case another CPU has just made room avail */ 3202 if (unlikely(dql_avail(&dev_queue->dql) >= 0)) 3203 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state); 3204 #endif 3205 } 3206 3207 /* Variant of netdev_tx_sent_queue() for drivers that are aware 3208 * that they should not test BQL status themselves. 3209 * We do want to change __QUEUE_STATE_STACK_XOFF only for the last 3210 * skb of a batch. 3211 * Returns true if the doorbell must be used to kick the NIC. 3212 */ 3213 static inline bool __netdev_tx_sent_queue(struct netdev_queue *dev_queue, 3214 unsigned int bytes, 3215 bool xmit_more) 3216 { 3217 if (xmit_more) { 3218 #ifdef CONFIG_BQL 3219 dql_queued(&dev_queue->dql, bytes); 3220 #endif 3221 return netif_tx_queue_stopped(dev_queue); 3222 } 3223 netdev_tx_sent_queue(dev_queue, bytes); 3224 return true; 3225 } 3226 3227 /** 3228 * netdev_sent_queue - report the number of bytes queued to hardware 3229 * @dev: network device 3230 * @bytes: number of bytes queued to the hardware device queue 3231 * 3232 * Report the number of bytes queued for sending/completion to the network 3233 * device hardware queue. @bytes should be a good approximation and should 3234 * exactly match netdev_completed_queue() @bytes 3235 */ 3236 static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes) 3237 { 3238 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes); 3239 } 3240 3241 static inline bool __netdev_sent_queue(struct net_device *dev, 3242 unsigned int bytes, 3243 bool xmit_more) 3244 { 3245 return __netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes, 3246 xmit_more); 3247 } 3248 3249 static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue, 3250 unsigned int pkts, unsigned int bytes) 3251 { 3252 #ifdef CONFIG_BQL 3253 if (unlikely(!bytes)) 3254 return; 3255 3256 dql_completed(&dev_queue->dql, bytes); 3257 3258 /* 3259 * Without the memory barrier there is a small possiblity that 3260 * netdev_tx_sent_queue will miss the update and cause the queue to 3261 * be stopped forever 3262 */ 3263 smp_mb(); 3264 3265 if (dql_avail(&dev_queue->dql) < 0) 3266 return; 3267 3268 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state)) 3269 netif_schedule_queue(dev_queue); 3270 #endif 3271 } 3272 3273 /** 3274 * netdev_completed_queue - report bytes and packets completed by device 3275 * @dev: network device 3276 * @pkts: actual number of packets sent over the medium 3277 * @bytes: actual number of bytes sent over the medium 3278 * 3279 * Report the number of bytes and packets transmitted by the network device 3280 * hardware queue over the physical medium, @bytes must exactly match the 3281 * @bytes amount passed to netdev_sent_queue() 3282 */ 3283 static inline void netdev_completed_queue(struct net_device *dev, 3284 unsigned int pkts, unsigned int bytes) 3285 { 3286 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes); 3287 } 3288 3289 static inline void netdev_tx_reset_queue(struct netdev_queue *q) 3290 { 3291 #ifdef CONFIG_BQL 3292 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state); 3293 dql_reset(&q->dql); 3294 #endif 3295 } 3296 3297 /** 3298 * netdev_reset_queue - reset the packets and bytes count of a network device 3299 * @dev_queue: network device 3300 * 3301 * Reset the bytes and packet count of a network device and clear the 3302 * software flow control OFF bit for this network device 3303 */ 3304 static inline void netdev_reset_queue(struct net_device *dev_queue) 3305 { 3306 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0)); 3307 } 3308 3309 /** 3310 * netdev_cap_txqueue - check if selected tx queue exceeds device queues 3311 * @dev: network device 3312 * @queue_index: given tx queue index 3313 * 3314 * Returns 0 if given tx queue index >= number of device tx queues, 3315 * otherwise returns the originally passed tx queue index. 3316 */ 3317 static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index) 3318 { 3319 if (unlikely(queue_index >= dev->real_num_tx_queues)) { 3320 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n", 3321 dev->name, queue_index, 3322 dev->real_num_tx_queues); 3323 return 0; 3324 } 3325 3326 return queue_index; 3327 } 3328 3329 /** 3330 * netif_running - test if up 3331 * @dev: network device 3332 * 3333 * Test if the device has been brought up. 3334 */ 3335 static inline bool netif_running(const struct net_device *dev) 3336 { 3337 return test_bit(__LINK_STATE_START, &dev->state); 3338 } 3339 3340 /* 3341 * Routines to manage the subqueues on a device. We only need start, 3342 * stop, and a check if it's stopped. All other device management is 3343 * done at the overall netdevice level. 3344 * Also test the device if we're multiqueue. 3345 */ 3346 3347 /** 3348 * netif_start_subqueue - allow sending packets on subqueue 3349 * @dev: network device 3350 * @queue_index: sub queue index 3351 * 3352 * Start individual transmit queue of a device with multiple transmit queues. 3353 */ 3354 static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index) 3355 { 3356 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 3357 3358 netif_tx_start_queue(txq); 3359 } 3360 3361 /** 3362 * netif_stop_subqueue - stop sending packets on subqueue 3363 * @dev: network device 3364 * @queue_index: sub queue index 3365 * 3366 * Stop individual transmit queue of a device with multiple transmit queues. 3367 */ 3368 static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index) 3369 { 3370 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 3371 netif_tx_stop_queue(txq); 3372 } 3373 3374 /** 3375 * netif_subqueue_stopped - test status of subqueue 3376 * @dev: network device 3377 * @queue_index: sub queue index 3378 * 3379 * Check individual transmit queue of a device with multiple transmit queues. 3380 */ 3381 static inline bool __netif_subqueue_stopped(const struct net_device *dev, 3382 u16 queue_index) 3383 { 3384 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 3385 3386 return netif_tx_queue_stopped(txq); 3387 } 3388 3389 static inline bool netif_subqueue_stopped(const struct net_device *dev, 3390 struct sk_buff *skb) 3391 { 3392 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb)); 3393 } 3394 3395 /** 3396 * netif_wake_subqueue - allow sending packets on subqueue 3397 * @dev: network device 3398 * @queue_index: sub queue index 3399 * 3400 * Resume individual transmit queue of a device with multiple transmit queues. 3401 */ 3402 static inline void netif_wake_subqueue(struct net_device *dev, u16 queue_index) 3403 { 3404 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 3405 3406 netif_tx_wake_queue(txq); 3407 } 3408 3409 #ifdef CONFIG_XPS 3410 int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask, 3411 u16 index); 3412 int __netif_set_xps_queue(struct net_device *dev, const unsigned long *mask, 3413 u16 index, bool is_rxqs_map); 3414 3415 /** 3416 * netif_attr_test_mask - Test a CPU or Rx queue set in a mask 3417 * @j: CPU/Rx queue index 3418 * @mask: bitmask of all cpus/rx queues 3419 * @nr_bits: number of bits in the bitmask 3420 * 3421 * Test if a CPU or Rx queue index is set in a mask of all CPU/Rx queues. 3422 */ 3423 static inline bool netif_attr_test_mask(unsigned long j, 3424 const unsigned long *mask, 3425 unsigned int nr_bits) 3426 { 3427 cpu_max_bits_warn(j, nr_bits); 3428 return test_bit(j, mask); 3429 } 3430 3431 /** 3432 * netif_attr_test_online - Test for online CPU/Rx queue 3433 * @j: CPU/Rx queue index 3434 * @online_mask: bitmask for CPUs/Rx queues that are online 3435 * @nr_bits: number of bits in the bitmask 3436 * 3437 * Returns true if a CPU/Rx queue is online. 3438 */ 3439 static inline bool netif_attr_test_online(unsigned long j, 3440 const unsigned long *online_mask, 3441 unsigned int nr_bits) 3442 { 3443 cpu_max_bits_warn(j, nr_bits); 3444 3445 if (online_mask) 3446 return test_bit(j, online_mask); 3447 3448 return (j < nr_bits); 3449 } 3450 3451 /** 3452 * netif_attrmask_next - get the next CPU/Rx queue in a cpu/Rx queues mask 3453 * @n: CPU/Rx queue index 3454 * @srcp: the cpumask/Rx queue mask pointer 3455 * @nr_bits: number of bits in the bitmask 3456 * 3457 * Returns >= nr_bits if no further CPUs/Rx queues set. 3458 */ 3459 static inline unsigned int netif_attrmask_next(int n, const unsigned long *srcp, 3460 unsigned int nr_bits) 3461 { 3462 /* -1 is a legal arg here. */ 3463 if (n != -1) 3464 cpu_max_bits_warn(n, nr_bits); 3465 3466 if (srcp) 3467 return find_next_bit(srcp, nr_bits, n + 1); 3468 3469 return n + 1; 3470 } 3471 3472 /** 3473 * netif_attrmask_next_and - get the next CPU/Rx queue in *src1p & *src2p 3474 * @n: CPU/Rx queue index 3475 * @src1p: the first CPUs/Rx queues mask pointer 3476 * @src2p: the second CPUs/Rx queues mask pointer 3477 * @nr_bits: number of bits in the bitmask 3478 * 3479 * Returns >= nr_bits if no further CPUs/Rx queues set in both. 3480 */ 3481 static inline int netif_attrmask_next_and(int n, const unsigned long *src1p, 3482 const unsigned long *src2p, 3483 unsigned int nr_bits) 3484 { 3485 /* -1 is a legal arg here. */ 3486 if (n != -1) 3487 cpu_max_bits_warn(n, nr_bits); 3488 3489 if (src1p && src2p) 3490 return find_next_and_bit(src1p, src2p, nr_bits, n + 1); 3491 else if (src1p) 3492 return find_next_bit(src1p, nr_bits, n + 1); 3493 else if (src2p) 3494 return find_next_bit(src2p, nr_bits, n + 1); 3495 3496 return n + 1; 3497 } 3498 #else 3499 static inline int netif_set_xps_queue(struct net_device *dev, 3500 const struct cpumask *mask, 3501 u16 index) 3502 { 3503 return 0; 3504 } 3505 3506 static inline int __netif_set_xps_queue(struct net_device *dev, 3507 const unsigned long *mask, 3508 u16 index, bool is_rxqs_map) 3509 { 3510 return 0; 3511 } 3512 #endif 3513 3514 /** 3515 * netif_is_multiqueue - test if device has multiple transmit queues 3516 * @dev: network device 3517 * 3518 * Check if device has multiple transmit queues 3519 */ 3520 static inline bool netif_is_multiqueue(const struct net_device *dev) 3521 { 3522 return dev->num_tx_queues > 1; 3523 } 3524 3525 int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq); 3526 3527 #ifdef CONFIG_SYSFS 3528 int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq); 3529 #else 3530 static inline int netif_set_real_num_rx_queues(struct net_device *dev, 3531 unsigned int rxqs) 3532 { 3533 dev->real_num_rx_queues = rxqs; 3534 return 0; 3535 } 3536 #endif 3537 3538 static inline struct netdev_rx_queue * 3539 __netif_get_rx_queue(struct net_device *dev, unsigned int rxq) 3540 { 3541 return dev->_rx + rxq; 3542 } 3543 3544 #ifdef CONFIG_SYSFS 3545 static inline unsigned int get_netdev_rx_queue_index( 3546 struct netdev_rx_queue *queue) 3547 { 3548 struct net_device *dev = queue->dev; 3549 int index = queue - dev->_rx; 3550 3551 BUG_ON(index >= dev->num_rx_queues); 3552 return index; 3553 } 3554 #endif 3555 3556 #define DEFAULT_MAX_NUM_RSS_QUEUES (8) 3557 int netif_get_num_default_rss_queues(void); 3558 3559 enum skb_free_reason { 3560 SKB_REASON_CONSUMED, 3561 SKB_REASON_DROPPED, 3562 }; 3563 3564 void __dev_kfree_skb_irq(struct sk_buff *skb, enum skb_free_reason reason); 3565 void __dev_kfree_skb_any(struct sk_buff *skb, enum skb_free_reason reason); 3566 3567 /* 3568 * It is not allowed to call kfree_skb() or consume_skb() from hardware 3569 * interrupt context or with hardware interrupts being disabled. 3570 * (in_irq() || irqs_disabled()) 3571 * 3572 * We provide four helpers that can be used in following contexts : 3573 * 3574 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context, 3575 * replacing kfree_skb(skb) 3576 * 3577 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context. 3578 * Typically used in place of consume_skb(skb) in TX completion path 3579 * 3580 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context, 3581 * replacing kfree_skb(skb) 3582 * 3583 * dev_consume_skb_any(skb) when caller doesn't know its current irq context, 3584 * and consumed a packet. Used in place of consume_skb(skb) 3585 */ 3586 static inline void dev_kfree_skb_irq(struct sk_buff *skb) 3587 { 3588 __dev_kfree_skb_irq(skb, SKB_REASON_DROPPED); 3589 } 3590 3591 static inline void dev_consume_skb_irq(struct sk_buff *skb) 3592 { 3593 __dev_kfree_skb_irq(skb, SKB_REASON_CONSUMED); 3594 } 3595 3596 static inline void dev_kfree_skb_any(struct sk_buff *skb) 3597 { 3598 __dev_kfree_skb_any(skb, SKB_REASON_DROPPED); 3599 } 3600 3601 static inline void dev_consume_skb_any(struct sk_buff *skb) 3602 { 3603 __dev_kfree_skb_any(skb, SKB_REASON_CONSUMED); 3604 } 3605 3606 void generic_xdp_tx(struct sk_buff *skb, struct bpf_prog *xdp_prog); 3607 int do_xdp_generic(struct bpf_prog *xdp_prog, struct sk_buff *skb); 3608 int netif_rx(struct sk_buff *skb); 3609 int netif_rx_ni(struct sk_buff *skb); 3610 int netif_receive_skb(struct sk_buff *skb); 3611 int netif_receive_skb_core(struct sk_buff *skb); 3612 void netif_receive_skb_list(struct list_head *head); 3613 gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb); 3614 void napi_gro_flush(struct napi_struct *napi, bool flush_old); 3615 struct sk_buff *napi_get_frags(struct napi_struct *napi); 3616 gro_result_t napi_gro_frags(struct napi_struct *napi); 3617 struct packet_offload *gro_find_receive_by_type(__be16 type); 3618 struct packet_offload *gro_find_complete_by_type(__be16 type); 3619 3620 static inline void napi_free_frags(struct napi_struct *napi) 3621 { 3622 kfree_skb(napi->skb); 3623 napi->skb = NULL; 3624 } 3625 3626 bool netdev_is_rx_handler_busy(struct net_device *dev); 3627 int netdev_rx_handler_register(struct net_device *dev, 3628 rx_handler_func_t *rx_handler, 3629 void *rx_handler_data); 3630 void netdev_rx_handler_unregister(struct net_device *dev); 3631 3632 bool dev_valid_name(const char *name); 3633 int dev_ioctl(struct net *net, unsigned int cmd, struct ifreq *ifr, 3634 bool *need_copyout); 3635 int dev_ifconf(struct net *net, struct ifconf *, int); 3636 int dev_ethtool(struct net *net, struct ifreq *); 3637 unsigned int dev_get_flags(const struct net_device *); 3638 int __dev_change_flags(struct net_device *dev, unsigned int flags, 3639 struct netlink_ext_ack *extack); 3640 int dev_change_flags(struct net_device *dev, unsigned int flags, 3641 struct netlink_ext_ack *extack); 3642 void __dev_notify_flags(struct net_device *, unsigned int old_flags, 3643 unsigned int gchanges); 3644 int dev_change_name(struct net_device *, const char *); 3645 int dev_set_alias(struct net_device *, const char *, size_t); 3646 int dev_get_alias(const struct net_device *, char *, size_t); 3647 int dev_change_net_namespace(struct net_device *, struct net *, const char *); 3648 int __dev_set_mtu(struct net_device *, int); 3649 int dev_set_mtu_ext(struct net_device *dev, int mtu, 3650 struct netlink_ext_ack *extack); 3651 int dev_set_mtu(struct net_device *, int); 3652 int dev_change_tx_queue_len(struct net_device *, unsigned long); 3653 void dev_set_group(struct net_device *, int); 3654 int dev_pre_changeaddr_notify(struct net_device *dev, const char *addr, 3655 struct netlink_ext_ack *extack); 3656 int dev_set_mac_address(struct net_device *dev, struct sockaddr *sa, 3657 struct netlink_ext_ack *extack); 3658 int dev_change_carrier(struct net_device *, bool new_carrier); 3659 int dev_get_phys_port_id(struct net_device *dev, 3660 struct netdev_phys_item_id *ppid); 3661 int dev_get_phys_port_name(struct net_device *dev, 3662 char *name, size_t len); 3663 int dev_get_port_parent_id(struct net_device *dev, 3664 struct netdev_phys_item_id *ppid, bool recurse); 3665 bool netdev_port_same_parent_id(struct net_device *a, struct net_device *b); 3666 int dev_change_proto_down(struct net_device *dev, bool proto_down); 3667 struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev, bool *again); 3668 struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev, 3669 struct netdev_queue *txq, int *ret); 3670 3671 typedef int (*bpf_op_t)(struct net_device *dev, struct netdev_bpf *bpf); 3672 int dev_change_xdp_fd(struct net_device *dev, struct netlink_ext_ack *extack, 3673 int fd, u32 flags); 3674 u32 __dev_xdp_query(struct net_device *dev, bpf_op_t xdp_op, 3675 enum bpf_netdev_command cmd); 3676 int xdp_umem_query(struct net_device *dev, u16 queue_id); 3677 3678 int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb); 3679 int dev_forward_skb(struct net_device *dev, struct sk_buff *skb); 3680 bool is_skb_forwardable(const struct net_device *dev, 3681 const struct sk_buff *skb); 3682 3683 static __always_inline int ____dev_forward_skb(struct net_device *dev, 3684 struct sk_buff *skb) 3685 { 3686 if (skb_orphan_frags(skb, GFP_ATOMIC) || 3687 unlikely(!is_skb_forwardable(dev, skb))) { 3688 atomic_long_inc(&dev->rx_dropped); 3689 kfree_skb(skb); 3690 return NET_RX_DROP; 3691 } 3692 3693 skb_scrub_packet(skb, true); 3694 skb->priority = 0; 3695 return 0; 3696 } 3697 3698 bool dev_nit_active(struct net_device *dev); 3699 void dev_queue_xmit_nit(struct sk_buff *skb, struct net_device *dev); 3700 3701 extern int netdev_budget; 3702 extern unsigned int netdev_budget_usecs; 3703 3704 /* Called by rtnetlink.c:rtnl_unlock() */ 3705 void netdev_run_todo(void); 3706 3707 /** 3708 * dev_put - release reference to device 3709 * @dev: network device 3710 * 3711 * Release reference to device to allow it to be freed. 3712 */ 3713 static inline void dev_put(struct net_device *dev) 3714 { 3715 this_cpu_dec(*dev->pcpu_refcnt); 3716 } 3717 3718 /** 3719 * dev_hold - get reference to device 3720 * @dev: network device 3721 * 3722 * Hold reference to device to keep it from being freed. 3723 */ 3724 static inline void dev_hold(struct net_device *dev) 3725 { 3726 this_cpu_inc(*dev->pcpu_refcnt); 3727 } 3728 3729 /* Carrier loss detection, dial on demand. The functions netif_carrier_on 3730 * and _off may be called from IRQ context, but it is caller 3731 * who is responsible for serialization of these calls. 3732 * 3733 * The name carrier is inappropriate, these functions should really be 3734 * called netif_lowerlayer_*() because they represent the state of any 3735 * kind of lower layer not just hardware media. 3736 */ 3737 3738 void linkwatch_init_dev(struct net_device *dev); 3739 void linkwatch_fire_event(struct net_device *dev); 3740 void linkwatch_forget_dev(struct net_device *dev); 3741 3742 /** 3743 * netif_carrier_ok - test if carrier present 3744 * @dev: network device 3745 * 3746 * Check if carrier is present on device 3747 */ 3748 static inline bool netif_carrier_ok(const struct net_device *dev) 3749 { 3750 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state); 3751 } 3752 3753 unsigned long dev_trans_start(struct net_device *dev); 3754 3755 void __netdev_watchdog_up(struct net_device *dev); 3756 3757 void netif_carrier_on(struct net_device *dev); 3758 3759 void netif_carrier_off(struct net_device *dev); 3760 3761 /** 3762 * netif_dormant_on - mark device as dormant. 3763 * @dev: network device 3764 * 3765 * Mark device as dormant (as per RFC2863). 3766 * 3767 * The dormant state indicates that the relevant interface is not 3768 * actually in a condition to pass packets (i.e., it is not 'up') but is 3769 * in a "pending" state, waiting for some external event. For "on- 3770 * demand" interfaces, this new state identifies the situation where the 3771 * interface is waiting for events to place it in the up state. 3772 */ 3773 static inline void netif_dormant_on(struct net_device *dev) 3774 { 3775 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state)) 3776 linkwatch_fire_event(dev); 3777 } 3778 3779 /** 3780 * netif_dormant_off - set device as not dormant. 3781 * @dev: network device 3782 * 3783 * Device is not in dormant state. 3784 */ 3785 static inline void netif_dormant_off(struct net_device *dev) 3786 { 3787 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state)) 3788 linkwatch_fire_event(dev); 3789 } 3790 3791 /** 3792 * netif_dormant - test if device is dormant 3793 * @dev: network device 3794 * 3795 * Check if device is dormant. 3796 */ 3797 static inline bool netif_dormant(const struct net_device *dev) 3798 { 3799 return test_bit(__LINK_STATE_DORMANT, &dev->state); 3800 } 3801 3802 3803 /** 3804 * netif_oper_up - test if device is operational 3805 * @dev: network device 3806 * 3807 * Check if carrier is operational 3808 */ 3809 static inline bool netif_oper_up(const struct net_device *dev) 3810 { 3811 return (dev->operstate == IF_OPER_UP || 3812 dev->operstate == IF_OPER_UNKNOWN /* backward compat */); 3813 } 3814 3815 /** 3816 * netif_device_present - is device available or removed 3817 * @dev: network device 3818 * 3819 * Check if device has not been removed from system. 3820 */ 3821 static inline bool netif_device_present(struct net_device *dev) 3822 { 3823 return test_bit(__LINK_STATE_PRESENT, &dev->state); 3824 } 3825 3826 void netif_device_detach(struct net_device *dev); 3827 3828 void netif_device_attach(struct net_device *dev); 3829 3830 /* 3831 * Network interface message level settings 3832 */ 3833 3834 enum { 3835 NETIF_MSG_DRV = 0x0001, 3836 NETIF_MSG_PROBE = 0x0002, 3837 NETIF_MSG_LINK = 0x0004, 3838 NETIF_MSG_TIMER = 0x0008, 3839 NETIF_MSG_IFDOWN = 0x0010, 3840 NETIF_MSG_IFUP = 0x0020, 3841 NETIF_MSG_RX_ERR = 0x0040, 3842 NETIF_MSG_TX_ERR = 0x0080, 3843 NETIF_MSG_TX_QUEUED = 0x0100, 3844 NETIF_MSG_INTR = 0x0200, 3845 NETIF_MSG_TX_DONE = 0x0400, 3846 NETIF_MSG_RX_STATUS = 0x0800, 3847 NETIF_MSG_PKTDATA = 0x1000, 3848 NETIF_MSG_HW = 0x2000, 3849 NETIF_MSG_WOL = 0x4000, 3850 }; 3851 3852 #define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV) 3853 #define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE) 3854 #define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK) 3855 #define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER) 3856 #define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN) 3857 #define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP) 3858 #define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR) 3859 #define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR) 3860 #define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED) 3861 #define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR) 3862 #define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE) 3863 #define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS) 3864 #define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA) 3865 #define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW) 3866 #define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL) 3867 3868 static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits) 3869 { 3870 /* use default */ 3871 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8)) 3872 return default_msg_enable_bits; 3873 if (debug_value == 0) /* no output */ 3874 return 0; 3875 /* set low N bits */ 3876 return (1 << debug_value) - 1; 3877 } 3878 3879 static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu) 3880 { 3881 spin_lock(&txq->_xmit_lock); 3882 txq->xmit_lock_owner = cpu; 3883 } 3884 3885 static inline bool __netif_tx_acquire(struct netdev_queue *txq) 3886 { 3887 __acquire(&txq->_xmit_lock); 3888 return true; 3889 } 3890 3891 static inline void __netif_tx_release(struct netdev_queue *txq) 3892 { 3893 __release(&txq->_xmit_lock); 3894 } 3895 3896 static inline void __netif_tx_lock_bh(struct netdev_queue *txq) 3897 { 3898 spin_lock_bh(&txq->_xmit_lock); 3899 txq->xmit_lock_owner = smp_processor_id(); 3900 } 3901 3902 static inline bool __netif_tx_trylock(struct netdev_queue *txq) 3903 { 3904 bool ok = spin_trylock(&txq->_xmit_lock); 3905 if (likely(ok)) 3906 txq->xmit_lock_owner = smp_processor_id(); 3907 return ok; 3908 } 3909 3910 static inline void __netif_tx_unlock(struct netdev_queue *txq) 3911 { 3912 txq->xmit_lock_owner = -1; 3913 spin_unlock(&txq->_xmit_lock); 3914 } 3915 3916 static inline void __netif_tx_unlock_bh(struct netdev_queue *txq) 3917 { 3918 txq->xmit_lock_owner = -1; 3919 spin_unlock_bh(&txq->_xmit_lock); 3920 } 3921 3922 static inline void txq_trans_update(struct netdev_queue *txq) 3923 { 3924 if (txq->xmit_lock_owner != -1) 3925 txq->trans_start = jiffies; 3926 } 3927 3928 /* legacy drivers only, netdev_start_xmit() sets txq->trans_start */ 3929 static inline void netif_trans_update(struct net_device *dev) 3930 { 3931 struct netdev_queue *txq = netdev_get_tx_queue(dev, 0); 3932 3933 if (txq->trans_start != jiffies) 3934 txq->trans_start = jiffies; 3935 } 3936 3937 /** 3938 * netif_tx_lock - grab network device transmit lock 3939 * @dev: network device 3940 * 3941 * Get network device transmit lock 3942 */ 3943 static inline void netif_tx_lock(struct net_device *dev) 3944 { 3945 unsigned int i; 3946 int cpu; 3947 3948 spin_lock(&dev->tx_global_lock); 3949 cpu = smp_processor_id(); 3950 for (i = 0; i < dev->num_tx_queues; i++) { 3951 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 3952 3953 /* We are the only thread of execution doing a 3954 * freeze, but we have to grab the _xmit_lock in 3955 * order to synchronize with threads which are in 3956 * the ->hard_start_xmit() handler and already 3957 * checked the frozen bit. 3958 */ 3959 __netif_tx_lock(txq, cpu); 3960 set_bit(__QUEUE_STATE_FROZEN, &txq->state); 3961 __netif_tx_unlock(txq); 3962 } 3963 } 3964 3965 static inline void netif_tx_lock_bh(struct net_device *dev) 3966 { 3967 local_bh_disable(); 3968 netif_tx_lock(dev); 3969 } 3970 3971 static inline void netif_tx_unlock(struct net_device *dev) 3972 { 3973 unsigned int i; 3974 3975 for (i = 0; i < dev->num_tx_queues; i++) { 3976 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 3977 3978 /* No need to grab the _xmit_lock here. If the 3979 * queue is not stopped for another reason, we 3980 * force a schedule. 3981 */ 3982 clear_bit(__QUEUE_STATE_FROZEN, &txq->state); 3983 netif_schedule_queue(txq); 3984 } 3985 spin_unlock(&dev->tx_global_lock); 3986 } 3987 3988 static inline void netif_tx_unlock_bh(struct net_device *dev) 3989 { 3990 netif_tx_unlock(dev); 3991 local_bh_enable(); 3992 } 3993 3994 #define HARD_TX_LOCK(dev, txq, cpu) { \ 3995 if ((dev->features & NETIF_F_LLTX) == 0) { \ 3996 __netif_tx_lock(txq, cpu); \ 3997 } else { \ 3998 __netif_tx_acquire(txq); \ 3999 } \ 4000 } 4001 4002 #define HARD_TX_TRYLOCK(dev, txq) \ 4003 (((dev->features & NETIF_F_LLTX) == 0) ? \ 4004 __netif_tx_trylock(txq) : \ 4005 __netif_tx_acquire(txq)) 4006 4007 #define HARD_TX_UNLOCK(dev, txq) { \ 4008 if ((dev->features & NETIF_F_LLTX) == 0) { \ 4009 __netif_tx_unlock(txq); \ 4010 } else { \ 4011 __netif_tx_release(txq); \ 4012 } \ 4013 } 4014 4015 static inline void netif_tx_disable(struct net_device *dev) 4016 { 4017 unsigned int i; 4018 int cpu; 4019 4020 local_bh_disable(); 4021 cpu = smp_processor_id(); 4022 for (i = 0; i < dev->num_tx_queues; i++) { 4023 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 4024 4025 __netif_tx_lock(txq, cpu); 4026 netif_tx_stop_queue(txq); 4027 __netif_tx_unlock(txq); 4028 } 4029 local_bh_enable(); 4030 } 4031 4032 static inline void netif_addr_lock(struct net_device *dev) 4033 { 4034 spin_lock(&dev->addr_list_lock); 4035 } 4036 4037 static inline void netif_addr_lock_nested(struct net_device *dev) 4038 { 4039 int subclass = SINGLE_DEPTH_NESTING; 4040 4041 if (dev->netdev_ops->ndo_get_lock_subclass) 4042 subclass = dev->netdev_ops->ndo_get_lock_subclass(dev); 4043 4044 spin_lock_nested(&dev->addr_list_lock, subclass); 4045 } 4046 4047 static inline void netif_addr_lock_bh(struct net_device *dev) 4048 { 4049 spin_lock_bh(&dev->addr_list_lock); 4050 } 4051 4052 static inline void netif_addr_unlock(struct net_device *dev) 4053 { 4054 spin_unlock(&dev->addr_list_lock); 4055 } 4056 4057 static inline void netif_addr_unlock_bh(struct net_device *dev) 4058 { 4059 spin_unlock_bh(&dev->addr_list_lock); 4060 } 4061 4062 /* 4063 * dev_addrs walker. Should be used only for read access. Call with 4064 * rcu_read_lock held. 4065 */ 4066 #define for_each_dev_addr(dev, ha) \ 4067 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list) 4068 4069 /* These functions live elsewhere (drivers/net/net_init.c, but related) */ 4070 4071 void ether_setup(struct net_device *dev); 4072 4073 /* Support for loadable net-drivers */ 4074 struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name, 4075 unsigned char name_assign_type, 4076 void (*setup)(struct net_device *), 4077 unsigned int txqs, unsigned int rxqs); 4078 int dev_get_valid_name(struct net *net, struct net_device *dev, 4079 const char *name); 4080 4081 #define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \ 4082 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1) 4083 4084 #define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \ 4085 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \ 4086 count) 4087 4088 int register_netdev(struct net_device *dev); 4089 void unregister_netdev(struct net_device *dev); 4090 4091 /* General hardware address lists handling functions */ 4092 int __hw_addr_sync(struct netdev_hw_addr_list *to_list, 4093 struct netdev_hw_addr_list *from_list, int addr_len); 4094 void __hw_addr_unsync(struct netdev_hw_addr_list *to_list, 4095 struct netdev_hw_addr_list *from_list, int addr_len); 4096 int __hw_addr_sync_dev(struct netdev_hw_addr_list *list, 4097 struct net_device *dev, 4098 int (*sync)(struct net_device *, const unsigned char *), 4099 int (*unsync)(struct net_device *, 4100 const unsigned char *)); 4101 int __hw_addr_ref_sync_dev(struct netdev_hw_addr_list *list, 4102 struct net_device *dev, 4103 int (*sync)(struct net_device *, 4104 const unsigned char *, int), 4105 int (*unsync)(struct net_device *, 4106 const unsigned char *, int)); 4107 void __hw_addr_ref_unsync_dev(struct netdev_hw_addr_list *list, 4108 struct net_device *dev, 4109 int (*unsync)(struct net_device *, 4110 const unsigned char *, int)); 4111 void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list, 4112 struct net_device *dev, 4113 int (*unsync)(struct net_device *, 4114 const unsigned char *)); 4115 void __hw_addr_init(struct netdev_hw_addr_list *list); 4116 4117 /* Functions used for device addresses handling */ 4118 int dev_addr_add(struct net_device *dev, const unsigned char *addr, 4119 unsigned char addr_type); 4120 int dev_addr_del(struct net_device *dev, const unsigned char *addr, 4121 unsigned char addr_type); 4122 void dev_addr_flush(struct net_device *dev); 4123 int dev_addr_init(struct net_device *dev); 4124 4125 /* Functions used for unicast addresses handling */ 4126 int dev_uc_add(struct net_device *dev, const unsigned char *addr); 4127 int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr); 4128 int dev_uc_del(struct net_device *dev, const unsigned char *addr); 4129 int dev_uc_sync(struct net_device *to, struct net_device *from); 4130 int dev_uc_sync_multiple(struct net_device *to, struct net_device *from); 4131 void dev_uc_unsync(struct net_device *to, struct net_device *from); 4132 void dev_uc_flush(struct net_device *dev); 4133 void dev_uc_init(struct net_device *dev); 4134 4135 /** 4136 * __dev_uc_sync - Synchonize device's unicast list 4137 * @dev: device to sync 4138 * @sync: function to call if address should be added 4139 * @unsync: function to call if address should be removed 4140 * 4141 * Add newly added addresses to the interface, and release 4142 * addresses that have been deleted. 4143 */ 4144 static inline int __dev_uc_sync(struct net_device *dev, 4145 int (*sync)(struct net_device *, 4146 const unsigned char *), 4147 int (*unsync)(struct net_device *, 4148 const unsigned char *)) 4149 { 4150 return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync); 4151 } 4152 4153 /** 4154 * __dev_uc_unsync - Remove synchronized addresses from device 4155 * @dev: device to sync 4156 * @unsync: function to call if address should be removed 4157 * 4158 * Remove all addresses that were added to the device by dev_uc_sync(). 4159 */ 4160 static inline void __dev_uc_unsync(struct net_device *dev, 4161 int (*unsync)(struct net_device *, 4162 const unsigned char *)) 4163 { 4164 __hw_addr_unsync_dev(&dev->uc, dev, unsync); 4165 } 4166 4167 /* Functions used for multicast addresses handling */ 4168 int dev_mc_add(struct net_device *dev, const unsigned char *addr); 4169 int dev_mc_add_global(struct net_device *dev, const unsigned char *addr); 4170 int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr); 4171 int dev_mc_del(struct net_device *dev, const unsigned char *addr); 4172 int dev_mc_del_global(struct net_device *dev, const unsigned char *addr); 4173 int dev_mc_sync(struct net_device *to, struct net_device *from); 4174 int dev_mc_sync_multiple(struct net_device *to, struct net_device *from); 4175 void dev_mc_unsync(struct net_device *to, struct net_device *from); 4176 void dev_mc_flush(struct net_device *dev); 4177 void dev_mc_init(struct net_device *dev); 4178 4179 /** 4180 * __dev_mc_sync - Synchonize device's multicast list 4181 * @dev: device to sync 4182 * @sync: function to call if address should be added 4183 * @unsync: function to call if address should be removed 4184 * 4185 * Add newly added addresses to the interface, and release 4186 * addresses that have been deleted. 4187 */ 4188 static inline int __dev_mc_sync(struct net_device *dev, 4189 int (*sync)(struct net_device *, 4190 const unsigned char *), 4191 int (*unsync)(struct net_device *, 4192 const unsigned char *)) 4193 { 4194 return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync); 4195 } 4196 4197 /** 4198 * __dev_mc_unsync - Remove synchronized addresses from device 4199 * @dev: device to sync 4200 * @unsync: function to call if address should be removed 4201 * 4202 * Remove all addresses that were added to the device by dev_mc_sync(). 4203 */ 4204 static inline void __dev_mc_unsync(struct net_device *dev, 4205 int (*unsync)(struct net_device *, 4206 const unsigned char *)) 4207 { 4208 __hw_addr_unsync_dev(&dev->mc, dev, unsync); 4209 } 4210 4211 /* Functions used for secondary unicast and multicast support */ 4212 void dev_set_rx_mode(struct net_device *dev); 4213 void __dev_set_rx_mode(struct net_device *dev); 4214 int dev_set_promiscuity(struct net_device *dev, int inc); 4215 int dev_set_allmulti(struct net_device *dev, int inc); 4216 void netdev_state_change(struct net_device *dev); 4217 void netdev_notify_peers(struct net_device *dev); 4218 void netdev_features_change(struct net_device *dev); 4219 /* Load a device via the kmod */ 4220 void dev_load(struct net *net, const char *name); 4221 struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev, 4222 struct rtnl_link_stats64 *storage); 4223 void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64, 4224 const struct net_device_stats *netdev_stats); 4225 4226 extern int netdev_max_backlog; 4227 extern int netdev_tstamp_prequeue; 4228 extern int weight_p; 4229 extern int dev_weight_rx_bias; 4230 extern int dev_weight_tx_bias; 4231 extern int dev_rx_weight; 4232 extern int dev_tx_weight; 4233 4234 bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev); 4235 struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev, 4236 struct list_head **iter); 4237 struct net_device *netdev_all_upper_get_next_dev_rcu(struct net_device *dev, 4238 struct list_head **iter); 4239 4240 /* iterate through upper list, must be called under RCU read lock */ 4241 #define netdev_for_each_upper_dev_rcu(dev, updev, iter) \ 4242 for (iter = &(dev)->adj_list.upper, \ 4243 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \ 4244 updev; \ 4245 updev = netdev_upper_get_next_dev_rcu(dev, &(iter))) 4246 4247 int netdev_walk_all_upper_dev_rcu(struct net_device *dev, 4248 int (*fn)(struct net_device *upper_dev, 4249 void *data), 4250 void *data); 4251 4252 bool netdev_has_upper_dev_all_rcu(struct net_device *dev, 4253 struct net_device *upper_dev); 4254 4255 bool netdev_has_any_upper_dev(struct net_device *dev); 4256 4257 void *netdev_lower_get_next_private(struct net_device *dev, 4258 struct list_head **iter); 4259 void *netdev_lower_get_next_private_rcu(struct net_device *dev, 4260 struct list_head **iter); 4261 4262 #define netdev_for_each_lower_private(dev, priv, iter) \ 4263 for (iter = (dev)->adj_list.lower.next, \ 4264 priv = netdev_lower_get_next_private(dev, &(iter)); \ 4265 priv; \ 4266 priv = netdev_lower_get_next_private(dev, &(iter))) 4267 4268 #define netdev_for_each_lower_private_rcu(dev, priv, iter) \ 4269 for (iter = &(dev)->adj_list.lower, \ 4270 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \ 4271 priv; \ 4272 priv = netdev_lower_get_next_private_rcu(dev, &(iter))) 4273 4274 void *netdev_lower_get_next(struct net_device *dev, 4275 struct list_head **iter); 4276 4277 #define netdev_for_each_lower_dev(dev, ldev, iter) \ 4278 for (iter = (dev)->adj_list.lower.next, \ 4279 ldev = netdev_lower_get_next(dev, &(iter)); \ 4280 ldev; \ 4281 ldev = netdev_lower_get_next(dev, &(iter))) 4282 4283 struct net_device *netdev_all_lower_get_next(struct net_device *dev, 4284 struct list_head **iter); 4285 struct net_device *netdev_all_lower_get_next_rcu(struct net_device *dev, 4286 struct list_head **iter); 4287 4288 int netdev_walk_all_lower_dev(struct net_device *dev, 4289 int (*fn)(struct net_device *lower_dev, 4290 void *data), 4291 void *data); 4292 int netdev_walk_all_lower_dev_rcu(struct net_device *dev, 4293 int (*fn)(struct net_device *lower_dev, 4294 void *data), 4295 void *data); 4296 4297 void *netdev_adjacent_get_private(struct list_head *adj_list); 4298 void *netdev_lower_get_first_private_rcu(struct net_device *dev); 4299 struct net_device *netdev_master_upper_dev_get(struct net_device *dev); 4300 struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev); 4301 int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev, 4302 struct netlink_ext_ack *extack); 4303 int netdev_master_upper_dev_link(struct net_device *dev, 4304 struct net_device *upper_dev, 4305 void *upper_priv, void *upper_info, 4306 struct netlink_ext_ack *extack); 4307 void netdev_upper_dev_unlink(struct net_device *dev, 4308 struct net_device *upper_dev); 4309 void netdev_adjacent_rename_links(struct net_device *dev, char *oldname); 4310 void *netdev_lower_dev_get_private(struct net_device *dev, 4311 struct net_device *lower_dev); 4312 void netdev_lower_state_changed(struct net_device *lower_dev, 4313 void *lower_state_info); 4314 4315 /* RSS keys are 40 or 52 bytes long */ 4316 #define NETDEV_RSS_KEY_LEN 52 4317 extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN] __read_mostly; 4318 void netdev_rss_key_fill(void *buffer, size_t len); 4319 4320 int dev_get_nest_level(struct net_device *dev); 4321 int skb_checksum_help(struct sk_buff *skb); 4322 int skb_crc32c_csum_help(struct sk_buff *skb); 4323 int skb_csum_hwoffload_help(struct sk_buff *skb, 4324 const netdev_features_t features); 4325 4326 struct sk_buff *__skb_gso_segment(struct sk_buff *skb, 4327 netdev_features_t features, bool tx_path); 4328 struct sk_buff *skb_mac_gso_segment(struct sk_buff *skb, 4329 netdev_features_t features); 4330 4331 struct netdev_bonding_info { 4332 ifslave slave; 4333 ifbond master; 4334 }; 4335 4336 struct netdev_notifier_bonding_info { 4337 struct netdev_notifier_info info; /* must be first */ 4338 struct netdev_bonding_info bonding_info; 4339 }; 4340 4341 void netdev_bonding_info_change(struct net_device *dev, 4342 struct netdev_bonding_info *bonding_info); 4343 4344 static inline 4345 struct sk_buff *skb_gso_segment(struct sk_buff *skb, netdev_features_t features) 4346 { 4347 return __skb_gso_segment(skb, features, true); 4348 } 4349 __be16 skb_network_protocol(struct sk_buff *skb, int *depth); 4350 4351 static inline bool can_checksum_protocol(netdev_features_t features, 4352 __be16 protocol) 4353 { 4354 if (protocol == htons(ETH_P_FCOE)) 4355 return !!(features & NETIF_F_FCOE_CRC); 4356 4357 /* Assume this is an IP checksum (not SCTP CRC) */ 4358 4359 if (features & NETIF_F_HW_CSUM) { 4360 /* Can checksum everything */ 4361 return true; 4362 } 4363 4364 switch (protocol) { 4365 case htons(ETH_P_IP): 4366 return !!(features & NETIF_F_IP_CSUM); 4367 case htons(ETH_P_IPV6): 4368 return !!(features & NETIF_F_IPV6_CSUM); 4369 default: 4370 return false; 4371 } 4372 } 4373 4374 #ifdef CONFIG_BUG 4375 void netdev_rx_csum_fault(struct net_device *dev, struct sk_buff *skb); 4376 #else 4377 static inline void netdev_rx_csum_fault(struct net_device *dev, 4378 struct sk_buff *skb) 4379 { 4380 } 4381 #endif 4382 /* rx skb timestamps */ 4383 void net_enable_timestamp(void); 4384 void net_disable_timestamp(void); 4385 4386 #ifdef CONFIG_PROC_FS 4387 int __init dev_proc_init(void); 4388 #else 4389 #define dev_proc_init() 0 4390 #endif 4391 4392 static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops, 4393 struct sk_buff *skb, struct net_device *dev, 4394 bool more) 4395 { 4396 skb->xmit_more = more ? 1 : 0; 4397 return ops->ndo_start_xmit(skb, dev); 4398 } 4399 4400 static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev, 4401 struct netdev_queue *txq, bool more) 4402 { 4403 const struct net_device_ops *ops = dev->netdev_ops; 4404 netdev_tx_t rc; 4405 4406 rc = __netdev_start_xmit(ops, skb, dev, more); 4407 if (rc == NETDEV_TX_OK) 4408 txq_trans_update(txq); 4409 4410 return rc; 4411 } 4412 4413 int netdev_class_create_file_ns(const struct class_attribute *class_attr, 4414 const void *ns); 4415 void netdev_class_remove_file_ns(const struct class_attribute *class_attr, 4416 const void *ns); 4417 4418 static inline int netdev_class_create_file(const struct class_attribute *class_attr) 4419 { 4420 return netdev_class_create_file_ns(class_attr, NULL); 4421 } 4422 4423 static inline void netdev_class_remove_file(const struct class_attribute *class_attr) 4424 { 4425 netdev_class_remove_file_ns(class_attr, NULL); 4426 } 4427 4428 extern const struct kobj_ns_type_operations net_ns_type_operations; 4429 4430 const char *netdev_drivername(const struct net_device *dev); 4431 4432 void linkwatch_run_queue(void); 4433 4434 static inline netdev_features_t netdev_intersect_features(netdev_features_t f1, 4435 netdev_features_t f2) 4436 { 4437 if ((f1 ^ f2) & NETIF_F_HW_CSUM) { 4438 if (f1 & NETIF_F_HW_CSUM) 4439 f1 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM); 4440 else 4441 f2 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM); 4442 } 4443 4444 return f1 & f2; 4445 } 4446 4447 static inline netdev_features_t netdev_get_wanted_features( 4448 struct net_device *dev) 4449 { 4450 return (dev->features & ~dev->hw_features) | dev->wanted_features; 4451 } 4452 netdev_features_t netdev_increment_features(netdev_features_t all, 4453 netdev_features_t one, netdev_features_t mask); 4454 4455 /* Allow TSO being used on stacked device : 4456 * Performing the GSO segmentation before last device 4457 * is a performance improvement. 4458 */ 4459 static inline netdev_features_t netdev_add_tso_features(netdev_features_t features, 4460 netdev_features_t mask) 4461 { 4462 return netdev_increment_features(features, NETIF_F_ALL_TSO, mask); 4463 } 4464 4465 int __netdev_update_features(struct net_device *dev); 4466 void netdev_update_features(struct net_device *dev); 4467 void netdev_change_features(struct net_device *dev); 4468 4469 void netif_stacked_transfer_operstate(const struct net_device *rootdev, 4470 struct net_device *dev); 4471 4472 netdev_features_t passthru_features_check(struct sk_buff *skb, 4473 struct net_device *dev, 4474 netdev_features_t features); 4475 netdev_features_t netif_skb_features(struct sk_buff *skb); 4476 4477 static inline bool net_gso_ok(netdev_features_t features, int gso_type) 4478 { 4479 netdev_features_t feature = (netdev_features_t)gso_type << NETIF_F_GSO_SHIFT; 4480 4481 /* check flags correspondence */ 4482 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT)); 4483 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT)); 4484 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT)); 4485 BUILD_BUG_ON(SKB_GSO_TCP_FIXEDID != (NETIF_F_TSO_MANGLEID >> NETIF_F_GSO_SHIFT)); 4486 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT)); 4487 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT)); 4488 BUILD_BUG_ON(SKB_GSO_GRE != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT)); 4489 BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT)); 4490 BUILD_BUG_ON(SKB_GSO_IPXIP4 != (NETIF_F_GSO_IPXIP4 >> NETIF_F_GSO_SHIFT)); 4491 BUILD_BUG_ON(SKB_GSO_IPXIP6 != (NETIF_F_GSO_IPXIP6 >> NETIF_F_GSO_SHIFT)); 4492 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT)); 4493 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT)); 4494 BUILD_BUG_ON(SKB_GSO_PARTIAL != (NETIF_F_GSO_PARTIAL >> NETIF_F_GSO_SHIFT)); 4495 BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT)); 4496 BUILD_BUG_ON(SKB_GSO_SCTP != (NETIF_F_GSO_SCTP >> NETIF_F_GSO_SHIFT)); 4497 BUILD_BUG_ON(SKB_GSO_ESP != (NETIF_F_GSO_ESP >> NETIF_F_GSO_SHIFT)); 4498 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_GSO_UDP >> NETIF_F_GSO_SHIFT)); 4499 BUILD_BUG_ON(SKB_GSO_UDP_L4 != (NETIF_F_GSO_UDP_L4 >> NETIF_F_GSO_SHIFT)); 4500 4501 return (features & feature) == feature; 4502 } 4503 4504 static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features) 4505 { 4506 return net_gso_ok(features, skb_shinfo(skb)->gso_type) && 4507 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST)); 4508 } 4509 4510 static inline bool netif_needs_gso(struct sk_buff *skb, 4511 netdev_features_t features) 4512 { 4513 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) || 4514 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) && 4515 (skb->ip_summed != CHECKSUM_UNNECESSARY))); 4516 } 4517 4518 static inline void netif_set_gso_max_size(struct net_device *dev, 4519 unsigned int size) 4520 { 4521 dev->gso_max_size = size; 4522 } 4523 4524 static inline void skb_gso_error_unwind(struct sk_buff *skb, __be16 protocol, 4525 int pulled_hlen, u16 mac_offset, 4526 int mac_len) 4527 { 4528 skb->protocol = protocol; 4529 skb->encapsulation = 1; 4530 skb_push(skb, pulled_hlen); 4531 skb_reset_transport_header(skb); 4532 skb->mac_header = mac_offset; 4533 skb->network_header = skb->mac_header + mac_len; 4534 skb->mac_len = mac_len; 4535 } 4536 4537 static inline bool netif_is_macsec(const struct net_device *dev) 4538 { 4539 return dev->priv_flags & IFF_MACSEC; 4540 } 4541 4542 static inline bool netif_is_macvlan(const struct net_device *dev) 4543 { 4544 return dev->priv_flags & IFF_MACVLAN; 4545 } 4546 4547 static inline bool netif_is_macvlan_port(const struct net_device *dev) 4548 { 4549 return dev->priv_flags & IFF_MACVLAN_PORT; 4550 } 4551 4552 static inline bool netif_is_bond_master(const struct net_device *dev) 4553 { 4554 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING; 4555 } 4556 4557 static inline bool netif_is_bond_slave(const struct net_device *dev) 4558 { 4559 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING; 4560 } 4561 4562 static inline bool netif_supports_nofcs(struct net_device *dev) 4563 { 4564 return dev->priv_flags & IFF_SUPP_NOFCS; 4565 } 4566 4567 static inline bool netif_has_l3_rx_handler(const struct net_device *dev) 4568 { 4569 return dev->priv_flags & IFF_L3MDEV_RX_HANDLER; 4570 } 4571 4572 static inline bool netif_is_l3_master(const struct net_device *dev) 4573 { 4574 return dev->priv_flags & IFF_L3MDEV_MASTER; 4575 } 4576 4577 static inline bool netif_is_l3_slave(const struct net_device *dev) 4578 { 4579 return dev->priv_flags & IFF_L3MDEV_SLAVE; 4580 } 4581 4582 static inline bool netif_is_bridge_master(const struct net_device *dev) 4583 { 4584 return dev->priv_flags & IFF_EBRIDGE; 4585 } 4586 4587 static inline bool netif_is_bridge_port(const struct net_device *dev) 4588 { 4589 return dev->priv_flags & IFF_BRIDGE_PORT; 4590 } 4591 4592 static inline bool netif_is_ovs_master(const struct net_device *dev) 4593 { 4594 return dev->priv_flags & IFF_OPENVSWITCH; 4595 } 4596 4597 static inline bool netif_is_ovs_port(const struct net_device *dev) 4598 { 4599 return dev->priv_flags & IFF_OVS_DATAPATH; 4600 } 4601 4602 static inline bool netif_is_team_master(const struct net_device *dev) 4603 { 4604 return dev->priv_flags & IFF_TEAM; 4605 } 4606 4607 static inline bool netif_is_team_port(const struct net_device *dev) 4608 { 4609 return dev->priv_flags & IFF_TEAM_PORT; 4610 } 4611 4612 static inline bool netif_is_lag_master(const struct net_device *dev) 4613 { 4614 return netif_is_bond_master(dev) || netif_is_team_master(dev); 4615 } 4616 4617 static inline bool netif_is_lag_port(const struct net_device *dev) 4618 { 4619 return netif_is_bond_slave(dev) || netif_is_team_port(dev); 4620 } 4621 4622 static inline bool netif_is_rxfh_configured(const struct net_device *dev) 4623 { 4624 return dev->priv_flags & IFF_RXFH_CONFIGURED; 4625 } 4626 4627 static inline bool netif_is_failover(const struct net_device *dev) 4628 { 4629 return dev->priv_flags & IFF_FAILOVER; 4630 } 4631 4632 static inline bool netif_is_failover_slave(const struct net_device *dev) 4633 { 4634 return dev->priv_flags & IFF_FAILOVER_SLAVE; 4635 } 4636 4637 /* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */ 4638 static inline void netif_keep_dst(struct net_device *dev) 4639 { 4640 dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM); 4641 } 4642 4643 /* return true if dev can't cope with mtu frames that need vlan tag insertion */ 4644 static inline bool netif_reduces_vlan_mtu(struct net_device *dev) 4645 { 4646 /* TODO: reserve and use an additional IFF bit, if we get more users */ 4647 return dev->priv_flags & IFF_MACSEC; 4648 } 4649 4650 extern struct pernet_operations __net_initdata loopback_net_ops; 4651 4652 /* Logging, debugging and troubleshooting/diagnostic helpers. */ 4653 4654 /* netdev_printk helpers, similar to dev_printk */ 4655 4656 static inline const char *netdev_name(const struct net_device *dev) 4657 { 4658 if (!dev->name[0] || strchr(dev->name, '%')) 4659 return "(unnamed net_device)"; 4660 return dev->name; 4661 } 4662 4663 static inline bool netdev_unregistering(const struct net_device *dev) 4664 { 4665 return dev->reg_state == NETREG_UNREGISTERING; 4666 } 4667 4668 static inline const char *netdev_reg_state(const struct net_device *dev) 4669 { 4670 switch (dev->reg_state) { 4671 case NETREG_UNINITIALIZED: return " (uninitialized)"; 4672 case NETREG_REGISTERED: return ""; 4673 case NETREG_UNREGISTERING: return " (unregistering)"; 4674 case NETREG_UNREGISTERED: return " (unregistered)"; 4675 case NETREG_RELEASED: return " (released)"; 4676 case NETREG_DUMMY: return " (dummy)"; 4677 } 4678 4679 WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, dev->reg_state); 4680 return " (unknown)"; 4681 } 4682 4683 __printf(3, 4) __cold 4684 void netdev_printk(const char *level, const struct net_device *dev, 4685 const char *format, ...); 4686 __printf(2, 3) __cold 4687 void netdev_emerg(const struct net_device *dev, const char *format, ...); 4688 __printf(2, 3) __cold 4689 void netdev_alert(const struct net_device *dev, const char *format, ...); 4690 __printf(2, 3) __cold 4691 void netdev_crit(const struct net_device *dev, const char *format, ...); 4692 __printf(2, 3) __cold 4693 void netdev_err(const struct net_device *dev, const char *format, ...); 4694 __printf(2, 3) __cold 4695 void netdev_warn(const struct net_device *dev, const char *format, ...); 4696 __printf(2, 3) __cold 4697 void netdev_notice(const struct net_device *dev, const char *format, ...); 4698 __printf(2, 3) __cold 4699 void netdev_info(const struct net_device *dev, const char *format, ...); 4700 4701 #define netdev_level_once(level, dev, fmt, ...) \ 4702 do { \ 4703 static bool __print_once __read_mostly; \ 4704 \ 4705 if (!__print_once) { \ 4706 __print_once = true; \ 4707 netdev_printk(level, dev, fmt, ##__VA_ARGS__); \ 4708 } \ 4709 } while (0) 4710 4711 #define netdev_emerg_once(dev, fmt, ...) \ 4712 netdev_level_once(KERN_EMERG, dev, fmt, ##__VA_ARGS__) 4713 #define netdev_alert_once(dev, fmt, ...) \ 4714 netdev_level_once(KERN_ALERT, dev, fmt, ##__VA_ARGS__) 4715 #define netdev_crit_once(dev, fmt, ...) \ 4716 netdev_level_once(KERN_CRIT, dev, fmt, ##__VA_ARGS__) 4717 #define netdev_err_once(dev, fmt, ...) \ 4718 netdev_level_once(KERN_ERR, dev, fmt, ##__VA_ARGS__) 4719 #define netdev_warn_once(dev, fmt, ...) \ 4720 netdev_level_once(KERN_WARNING, dev, fmt, ##__VA_ARGS__) 4721 #define netdev_notice_once(dev, fmt, ...) \ 4722 netdev_level_once(KERN_NOTICE, dev, fmt, ##__VA_ARGS__) 4723 #define netdev_info_once(dev, fmt, ...) \ 4724 netdev_level_once(KERN_INFO, dev, fmt, ##__VA_ARGS__) 4725 4726 #define MODULE_ALIAS_NETDEV(device) \ 4727 MODULE_ALIAS("netdev-" device) 4728 4729 #if defined(CONFIG_DYNAMIC_DEBUG) 4730 #define netdev_dbg(__dev, format, args...) \ 4731 do { \ 4732 dynamic_netdev_dbg(__dev, format, ##args); \ 4733 } while (0) 4734 #elif defined(DEBUG) 4735 #define netdev_dbg(__dev, format, args...) \ 4736 netdev_printk(KERN_DEBUG, __dev, format, ##args) 4737 #else 4738 #define netdev_dbg(__dev, format, args...) \ 4739 ({ \ 4740 if (0) \ 4741 netdev_printk(KERN_DEBUG, __dev, format, ##args); \ 4742 }) 4743 #endif 4744 4745 #if defined(VERBOSE_DEBUG) 4746 #define netdev_vdbg netdev_dbg 4747 #else 4748 4749 #define netdev_vdbg(dev, format, args...) \ 4750 ({ \ 4751 if (0) \ 4752 netdev_printk(KERN_DEBUG, dev, format, ##args); \ 4753 0; \ 4754 }) 4755 #endif 4756 4757 /* 4758 * netdev_WARN() acts like dev_printk(), but with the key difference 4759 * of using a WARN/WARN_ON to get the message out, including the 4760 * file/line information and a backtrace. 4761 */ 4762 #define netdev_WARN(dev, format, args...) \ 4763 WARN(1, "netdevice: %s%s: " format, netdev_name(dev), \ 4764 netdev_reg_state(dev), ##args) 4765 4766 #define netdev_WARN_ONCE(dev, format, args...) \ 4767 WARN_ONCE(1, "netdevice: %s%s: " format, netdev_name(dev), \ 4768 netdev_reg_state(dev), ##args) 4769 4770 /* netif printk helpers, similar to netdev_printk */ 4771 4772 #define netif_printk(priv, type, level, dev, fmt, args...) \ 4773 do { \ 4774 if (netif_msg_##type(priv)) \ 4775 netdev_printk(level, (dev), fmt, ##args); \ 4776 } while (0) 4777 4778 #define netif_level(level, priv, type, dev, fmt, args...) \ 4779 do { \ 4780 if (netif_msg_##type(priv)) \ 4781 netdev_##level(dev, fmt, ##args); \ 4782 } while (0) 4783 4784 #define netif_emerg(priv, type, dev, fmt, args...) \ 4785 netif_level(emerg, priv, type, dev, fmt, ##args) 4786 #define netif_alert(priv, type, dev, fmt, args...) \ 4787 netif_level(alert, priv, type, dev, fmt, ##args) 4788 #define netif_crit(priv, type, dev, fmt, args...) \ 4789 netif_level(crit, priv, type, dev, fmt, ##args) 4790 #define netif_err(priv, type, dev, fmt, args...) \ 4791 netif_level(err, priv, type, dev, fmt, ##args) 4792 #define netif_warn(priv, type, dev, fmt, args...) \ 4793 netif_level(warn, priv, type, dev, fmt, ##args) 4794 #define netif_notice(priv, type, dev, fmt, args...) \ 4795 netif_level(notice, priv, type, dev, fmt, ##args) 4796 #define netif_info(priv, type, dev, fmt, args...) \ 4797 netif_level(info, priv, type, dev, fmt, ##args) 4798 4799 #if defined(CONFIG_DYNAMIC_DEBUG) 4800 #define netif_dbg(priv, type, netdev, format, args...) \ 4801 do { \ 4802 if (netif_msg_##type(priv)) \ 4803 dynamic_netdev_dbg(netdev, format, ##args); \ 4804 } while (0) 4805 #elif defined(DEBUG) 4806 #define netif_dbg(priv, type, dev, format, args...) \ 4807 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args) 4808 #else 4809 #define netif_dbg(priv, type, dev, format, args...) \ 4810 ({ \ 4811 if (0) \ 4812 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \ 4813 0; \ 4814 }) 4815 #endif 4816 4817 /* if @cond then downgrade to debug, else print at @level */ 4818 #define netif_cond_dbg(priv, type, netdev, cond, level, fmt, args...) \ 4819 do { \ 4820 if (cond) \ 4821 netif_dbg(priv, type, netdev, fmt, ##args); \ 4822 else \ 4823 netif_ ## level(priv, type, netdev, fmt, ##args); \ 4824 } while (0) 4825 4826 #if defined(VERBOSE_DEBUG) 4827 #define netif_vdbg netif_dbg 4828 #else 4829 #define netif_vdbg(priv, type, dev, format, args...) \ 4830 ({ \ 4831 if (0) \ 4832 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \ 4833 0; \ 4834 }) 4835 #endif 4836 4837 /* 4838 * The list of packet types we will receive (as opposed to discard) 4839 * and the routines to invoke. 4840 * 4841 * Why 16. Because with 16 the only overlap we get on a hash of the 4842 * low nibble of the protocol value is RARP/SNAP/X.25. 4843 * 4844 * 0800 IP 4845 * 0001 802.3 4846 * 0002 AX.25 4847 * 0004 802.2 4848 * 8035 RARP 4849 * 0005 SNAP 4850 * 0805 X.25 4851 * 0806 ARP 4852 * 8137 IPX 4853 * 0009 Localtalk 4854 * 86DD IPv6 4855 */ 4856 #define PTYPE_HASH_SIZE (16) 4857 #define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1) 4858 4859 #endif /* _LINUX_NETDEVICE_H */ 4860