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