1 /* 2 * INET An implementation of the TCP/IP protocol suite for the LINUX 3 * operating system. INET is implemented using the BSD Socket 4 * interface as the means of communication with the user level. 5 * 6 * Definitions for the Interfaces handler. 7 * 8 * Version: @(#)dev.h 1.0.10 08/12/93 9 * 10 * Authors: Ross Biro 11 * Fred N. van Kempen, <[email protected]> 12 * Corey Minyard <[email protected]> 13 * Donald J. Becker, <[email protected]> 14 * Alan Cox, <[email protected]> 15 * Bjorn Ekwall. <[email protected]> 16 * Pekka Riikonen <[email protected]> 17 * 18 * This program is free software; you can redistribute it and/or 19 * modify it under the terms of the GNU General Public License 20 * as published by the Free Software Foundation; either version 21 * 2 of the License, or (at your option) any later version. 22 * 23 * Moved to /usr/include/linux for NET3 24 */ 25 #ifndef _LINUX_NETDEVICE_H 26 #define _LINUX_NETDEVICE_H 27 28 #include <linux/timer.h> 29 #include <linux/bug.h> 30 #include <linux/delay.h> 31 #include <linux/atomic.h> 32 #include <linux/prefetch.h> 33 #include <asm/cache.h> 34 #include <asm/byteorder.h> 35 36 #include <linux/percpu.h> 37 #include <linux/rculist.h> 38 #include <linux/workqueue.h> 39 #include <linux/dynamic_queue_limits.h> 40 41 #include <linux/ethtool.h> 42 #include <net/net_namespace.h> 43 #ifdef CONFIG_DCB 44 #include <net/dcbnl.h> 45 #endif 46 #include <net/netprio_cgroup.h> 47 #include <net/xdp.h> 48 49 #include <linux/netdev_features.h> 50 #include <linux/neighbour.h> 51 #include <uapi/linux/netdevice.h> 52 #include <uapi/linux/if_bonding.h> 53 #include <uapi/linux/pkt_cls.h> 54 #include <linux/hashtable.h> 55 56 struct netpoll_info; 57 struct device; 58 struct phy_device; 59 struct dsa_port; 60 61 struct sfp_bus; 62 /* 802.11 specific */ 63 struct wireless_dev; 64 /* 802.15.4 specific */ 65 struct wpan_dev; 66 struct mpls_dev; 67 /* UDP Tunnel offloads */ 68 struct udp_tunnel_info; 69 struct bpf_prog; 70 struct xdp_buff; 71 72 void netdev_set_default_ethtool_ops(struct net_device *dev, 73 const struct ethtool_ops *ops); 74 75 /* Backlog congestion levels */ 76 #define NET_RX_SUCCESS 0 /* keep 'em coming, baby */ 77 #define NET_RX_DROP 1 /* packet dropped */ 78 79 /* 80 * Transmit return codes: transmit return codes originate from three different 81 * namespaces: 82 * 83 * - qdisc return codes 84 * - driver transmit return codes 85 * - errno values 86 * 87 * Drivers are allowed to return any one of those in their hard_start_xmit() 88 * function. Real network devices commonly used with qdiscs should only return 89 * the driver transmit return codes though - when qdiscs are used, the actual 90 * transmission happens asynchronously, so the value is not propagated to 91 * higher layers. Virtual network devices transmit synchronously; in this case 92 * the driver transmit return codes are consumed by dev_queue_xmit(), and all 93 * others are propagated to higher layers. 94 */ 95 96 /* qdisc ->enqueue() return codes. */ 97 #define NET_XMIT_SUCCESS 0x00 98 #define NET_XMIT_DROP 0x01 /* skb dropped */ 99 #define NET_XMIT_CN 0x02 /* congestion notification */ 100 #define NET_XMIT_MASK 0x0f /* qdisc flags in net/sch_generic.h */ 101 102 /* NET_XMIT_CN is special. It does not guarantee that this packet is lost. It 103 * indicates that the device will soon be dropping packets, or already drops 104 * some packets of the same priority; prompting us to send less aggressively. */ 105 #define net_xmit_eval(e) ((e) == NET_XMIT_CN ? 0 : (e)) 106 #define net_xmit_errno(e) ((e) != NET_XMIT_CN ? -ENOBUFS : 0) 107 108 /* Driver transmit return codes */ 109 #define NETDEV_TX_MASK 0xf0 110 111 enum netdev_tx { 112 __NETDEV_TX_MIN = INT_MIN, /* make sure enum is signed */ 113 NETDEV_TX_OK = 0x00, /* driver took care of packet */ 114 NETDEV_TX_BUSY = 0x10, /* driver tx path was busy*/ 115 }; 116 typedef enum netdev_tx netdev_tx_t; 117 118 /* 119 * Current order: NETDEV_TX_MASK > NET_XMIT_MASK >= 0 is significant; 120 * hard_start_xmit() return < NET_XMIT_MASK means skb was consumed. 121 */ 122 static inline bool dev_xmit_complete(int rc) 123 { 124 /* 125 * Positive cases with an skb consumed by a driver: 126 * - successful transmission (rc == NETDEV_TX_OK) 127 * - error while transmitting (rc < 0) 128 * - error while queueing to a different device (rc & NET_XMIT_MASK) 129 */ 130 if (likely(rc < NET_XMIT_MASK)) 131 return true; 132 133 return false; 134 } 135 136 /* 137 * Compute the worst-case header length according to the protocols 138 * used. 139 */ 140 141 #if defined(CONFIG_HYPERV_NET) 142 # define LL_MAX_HEADER 128 143 #elif defined(CONFIG_WLAN) || IS_ENABLED(CONFIG_AX25) 144 # if defined(CONFIG_MAC80211_MESH) 145 # define LL_MAX_HEADER 128 146 # else 147 # define LL_MAX_HEADER 96 148 # endif 149 #else 150 # define LL_MAX_HEADER 32 151 #endif 152 153 #if !IS_ENABLED(CONFIG_NET_IPIP) && !IS_ENABLED(CONFIG_NET_IPGRE) && \ 154 !IS_ENABLED(CONFIG_IPV6_SIT) && !IS_ENABLED(CONFIG_IPV6_TUNNEL) 155 #define MAX_HEADER LL_MAX_HEADER 156 #else 157 #define MAX_HEADER (LL_MAX_HEADER + 48) 158 #endif 159 160 /* 161 * Old network device statistics. Fields are native words 162 * (unsigned long) so they can be read and written atomically. 163 */ 164 165 struct net_device_stats { 166 unsigned long rx_packets; 167 unsigned long tx_packets; 168 unsigned long rx_bytes; 169 unsigned long tx_bytes; 170 unsigned long rx_errors; 171 unsigned long tx_errors; 172 unsigned long rx_dropped; 173 unsigned long tx_dropped; 174 unsigned long multicast; 175 unsigned long collisions; 176 unsigned long rx_length_errors; 177 unsigned long rx_over_errors; 178 unsigned long rx_crc_errors; 179 unsigned long rx_frame_errors; 180 unsigned long rx_fifo_errors; 181 unsigned long rx_missed_errors; 182 unsigned long tx_aborted_errors; 183 unsigned long tx_carrier_errors; 184 unsigned long tx_fifo_errors; 185 unsigned long tx_heartbeat_errors; 186 unsigned long tx_window_errors; 187 unsigned long rx_compressed; 188 unsigned long tx_compressed; 189 }; 190 191 192 #include <linux/cache.h> 193 #include <linux/skbuff.h> 194 195 #ifdef CONFIG_RPS 196 #include <linux/static_key.h> 197 extern struct static_key rps_needed; 198 extern struct static_key rfs_needed; 199 #endif 200 201 struct neighbour; 202 struct neigh_parms; 203 struct sk_buff; 204 205 struct netdev_hw_addr { 206 struct list_head list; 207 unsigned char addr[MAX_ADDR_LEN]; 208 unsigned char type; 209 #define NETDEV_HW_ADDR_T_LAN 1 210 #define NETDEV_HW_ADDR_T_SAN 2 211 #define NETDEV_HW_ADDR_T_SLAVE 3 212 #define NETDEV_HW_ADDR_T_UNICAST 4 213 #define NETDEV_HW_ADDR_T_MULTICAST 5 214 bool global_use; 215 int sync_cnt; 216 int refcount; 217 int synced; 218 struct rcu_head rcu_head; 219 }; 220 221 struct netdev_hw_addr_list { 222 struct list_head list; 223 int count; 224 }; 225 226 #define netdev_hw_addr_list_count(l) ((l)->count) 227 #define netdev_hw_addr_list_empty(l) (netdev_hw_addr_list_count(l) == 0) 228 #define netdev_hw_addr_list_for_each(ha, l) \ 229 list_for_each_entry(ha, &(l)->list, list) 230 231 #define netdev_uc_count(dev) netdev_hw_addr_list_count(&(dev)->uc) 232 #define netdev_uc_empty(dev) netdev_hw_addr_list_empty(&(dev)->uc) 233 #define netdev_for_each_uc_addr(ha, dev) \ 234 netdev_hw_addr_list_for_each(ha, &(dev)->uc) 235 236 #define netdev_mc_count(dev) netdev_hw_addr_list_count(&(dev)->mc) 237 #define netdev_mc_empty(dev) netdev_hw_addr_list_empty(&(dev)->mc) 238 #define netdev_for_each_mc_addr(ha, dev) \ 239 netdev_hw_addr_list_for_each(ha, &(dev)->mc) 240 241 struct hh_cache { 242 unsigned int hh_len; 243 seqlock_t hh_lock; 244 245 /* cached hardware header; allow for machine alignment needs. */ 246 #define HH_DATA_MOD 16 247 #define HH_DATA_OFF(__len) \ 248 (HH_DATA_MOD - (((__len - 1) & (HH_DATA_MOD - 1)) + 1)) 249 #define HH_DATA_ALIGN(__len) \ 250 (((__len)+(HH_DATA_MOD-1))&~(HH_DATA_MOD - 1)) 251 unsigned long hh_data[HH_DATA_ALIGN(LL_MAX_HEADER) / sizeof(long)]; 252 }; 253 254 /* Reserve HH_DATA_MOD byte-aligned hard_header_len, but at least that much. 255 * Alternative is: 256 * dev->hard_header_len ? (dev->hard_header_len + 257 * (HH_DATA_MOD - 1)) & ~(HH_DATA_MOD - 1) : 0 258 * 259 * We could use other alignment values, but we must maintain the 260 * relationship HH alignment <= LL alignment. 261 */ 262 #define LL_RESERVED_SPACE(dev) \ 263 ((((dev)->hard_header_len+(dev)->needed_headroom)&~(HH_DATA_MOD - 1)) + HH_DATA_MOD) 264 #define LL_RESERVED_SPACE_EXTRA(dev,extra) \ 265 ((((dev)->hard_header_len+(dev)->needed_headroom+(extra))&~(HH_DATA_MOD - 1)) + HH_DATA_MOD) 266 267 struct header_ops { 268 int (*create) (struct sk_buff *skb, struct net_device *dev, 269 unsigned short type, const void *daddr, 270 const void *saddr, unsigned int len); 271 int (*parse)(const struct sk_buff *skb, unsigned char *haddr); 272 int (*cache)(const struct neighbour *neigh, struct hh_cache *hh, __be16 type); 273 void (*cache_update)(struct hh_cache *hh, 274 const struct net_device *dev, 275 const unsigned char *haddr); 276 bool (*validate)(const char *ll_header, unsigned int len); 277 }; 278 279 /* These flag bits are private to the generic network queueing 280 * layer; they may not be explicitly referenced by any other 281 * code. 282 */ 283 284 enum netdev_state_t { 285 __LINK_STATE_START, 286 __LINK_STATE_PRESENT, 287 __LINK_STATE_NOCARRIER, 288 __LINK_STATE_LINKWATCH_PENDING, 289 __LINK_STATE_DORMANT, 290 }; 291 292 293 /* 294 * This structure holds boot-time configured netdevice settings. They 295 * are then used in the device probing. 296 */ 297 struct netdev_boot_setup { 298 char name[IFNAMSIZ]; 299 struct ifmap map; 300 }; 301 #define NETDEV_BOOT_SETUP_MAX 8 302 303 int __init netdev_boot_setup(char *str); 304 305 /* 306 * Structure for NAPI scheduling similar to tasklet but with weighting 307 */ 308 #define GRO_HASH_BUCKETS 8 309 struct napi_struct { 310 /* The poll_list must only be managed by the entity which 311 * changes the state of the NAPI_STATE_SCHED bit. This means 312 * whoever atomically sets that bit can add this napi_struct 313 * to the per-CPU poll_list, and whoever clears that bit 314 * can remove from the list right before clearing the bit. 315 */ 316 struct list_head poll_list; 317 318 unsigned long state; 319 int weight; 320 unsigned int gro_count; 321 int (*poll)(struct napi_struct *, int); 322 #ifdef CONFIG_NETPOLL 323 int poll_owner; 324 #endif 325 struct net_device *dev; 326 struct list_head gro_hash[GRO_HASH_BUCKETS]; 327 struct sk_buff *skb; 328 struct hrtimer timer; 329 struct list_head dev_list; 330 struct hlist_node napi_hash_node; 331 unsigned int napi_id; 332 }; 333 334 enum { 335 NAPI_STATE_SCHED, /* Poll is scheduled */ 336 NAPI_STATE_MISSED, /* reschedule a napi */ 337 NAPI_STATE_DISABLE, /* Disable pending */ 338 NAPI_STATE_NPSVC, /* Netpoll - don't dequeue from poll_list */ 339 NAPI_STATE_HASHED, /* In NAPI hash (busy polling possible) */ 340 NAPI_STATE_NO_BUSY_POLL,/* Do not add in napi_hash, no busy polling */ 341 NAPI_STATE_IN_BUSY_POLL,/* sk_busy_loop() owns this NAPI */ 342 }; 343 344 enum { 345 NAPIF_STATE_SCHED = BIT(NAPI_STATE_SCHED), 346 NAPIF_STATE_MISSED = BIT(NAPI_STATE_MISSED), 347 NAPIF_STATE_DISABLE = BIT(NAPI_STATE_DISABLE), 348 NAPIF_STATE_NPSVC = BIT(NAPI_STATE_NPSVC), 349 NAPIF_STATE_HASHED = BIT(NAPI_STATE_HASHED), 350 NAPIF_STATE_NO_BUSY_POLL = BIT(NAPI_STATE_NO_BUSY_POLL), 351 NAPIF_STATE_IN_BUSY_POLL = BIT(NAPI_STATE_IN_BUSY_POLL), 352 }; 353 354 enum gro_result { 355 GRO_MERGED, 356 GRO_MERGED_FREE, 357 GRO_HELD, 358 GRO_NORMAL, 359 GRO_DROP, 360 GRO_CONSUMED, 361 }; 362 typedef enum gro_result gro_result_t; 363 364 /* 365 * enum rx_handler_result - Possible return values for rx_handlers. 366 * @RX_HANDLER_CONSUMED: skb was consumed by rx_handler, do not process it 367 * further. 368 * @RX_HANDLER_ANOTHER: Do another round in receive path. This is indicated in 369 * case skb->dev was changed by rx_handler. 370 * @RX_HANDLER_EXACT: Force exact delivery, no wildcard. 371 * @RX_HANDLER_PASS: Do nothing, pass the skb as if no rx_handler was called. 372 * 373 * rx_handlers are functions called from inside __netif_receive_skb(), to do 374 * special processing of the skb, prior to delivery to protocol handlers. 375 * 376 * Currently, a net_device can only have a single rx_handler registered. Trying 377 * to register a second rx_handler will return -EBUSY. 378 * 379 * To register a rx_handler on a net_device, use netdev_rx_handler_register(). 380 * To unregister a rx_handler on a net_device, use 381 * netdev_rx_handler_unregister(). 382 * 383 * Upon return, rx_handler is expected to tell __netif_receive_skb() what to 384 * do with the skb. 385 * 386 * If the rx_handler consumed the skb in some way, it should return 387 * RX_HANDLER_CONSUMED. This is appropriate when the rx_handler arranged for 388 * the skb to be delivered in some other way. 389 * 390 * If the rx_handler changed skb->dev, to divert the skb to another 391 * net_device, it should return RX_HANDLER_ANOTHER. The rx_handler for the 392 * new device will be called if it exists. 393 * 394 * If the rx_handler decides the skb should be ignored, it should return 395 * RX_HANDLER_EXACT. The skb will only be delivered to protocol handlers that 396 * are registered on exact device (ptype->dev == skb->dev). 397 * 398 * If the rx_handler didn't change skb->dev, but wants the skb to be normally 399 * delivered, it should return RX_HANDLER_PASS. 400 * 401 * A device without a registered rx_handler will behave as if rx_handler 402 * returned RX_HANDLER_PASS. 403 */ 404 405 enum rx_handler_result { 406 RX_HANDLER_CONSUMED, 407 RX_HANDLER_ANOTHER, 408 RX_HANDLER_EXACT, 409 RX_HANDLER_PASS, 410 }; 411 typedef enum rx_handler_result rx_handler_result_t; 412 typedef rx_handler_result_t rx_handler_func_t(struct sk_buff **pskb); 413 414 void __napi_schedule(struct napi_struct *n); 415 void __napi_schedule_irqoff(struct napi_struct *n); 416 417 static inline bool napi_disable_pending(struct napi_struct *n) 418 { 419 return test_bit(NAPI_STATE_DISABLE, &n->state); 420 } 421 422 bool napi_schedule_prep(struct napi_struct *n); 423 424 /** 425 * napi_schedule - schedule NAPI poll 426 * @n: NAPI context 427 * 428 * Schedule NAPI poll routine to be called if it is not already 429 * running. 430 */ 431 static inline void napi_schedule(struct napi_struct *n) 432 { 433 if (napi_schedule_prep(n)) 434 __napi_schedule(n); 435 } 436 437 /** 438 * napi_schedule_irqoff - schedule NAPI poll 439 * @n: NAPI context 440 * 441 * Variant of napi_schedule(), assuming hard irqs are masked. 442 */ 443 static inline void napi_schedule_irqoff(struct napi_struct *n) 444 { 445 if (napi_schedule_prep(n)) 446 __napi_schedule_irqoff(n); 447 } 448 449 /* Try to reschedule poll. Called by dev->poll() after napi_complete(). */ 450 static inline bool napi_reschedule(struct napi_struct *napi) 451 { 452 if (napi_schedule_prep(napi)) { 453 __napi_schedule(napi); 454 return true; 455 } 456 return false; 457 } 458 459 bool napi_complete_done(struct napi_struct *n, int work_done); 460 /** 461 * napi_complete - NAPI processing complete 462 * @n: NAPI context 463 * 464 * Mark NAPI processing as complete. 465 * Consider using napi_complete_done() instead. 466 * Return false if device should avoid rearming interrupts. 467 */ 468 static inline bool napi_complete(struct napi_struct *n) 469 { 470 return napi_complete_done(n, 0); 471 } 472 473 /** 474 * napi_hash_del - remove a NAPI from global table 475 * @napi: NAPI context 476 * 477 * Warning: caller must observe RCU grace period 478 * before freeing memory containing @napi, if 479 * this function returns true. 480 * Note: core networking stack automatically calls it 481 * from netif_napi_del(). 482 * Drivers might want to call this helper to combine all 483 * the needed RCU grace periods into a single one. 484 */ 485 bool napi_hash_del(struct napi_struct *napi); 486 487 /** 488 * napi_disable - prevent NAPI from scheduling 489 * @n: NAPI context 490 * 491 * Stop NAPI from being scheduled on this context. 492 * Waits till any outstanding processing completes. 493 */ 494 void napi_disable(struct napi_struct *n); 495 496 /** 497 * napi_enable - enable NAPI scheduling 498 * @n: NAPI context 499 * 500 * Resume NAPI from being scheduled on this context. 501 * Must be paired with napi_disable. 502 */ 503 static inline void napi_enable(struct napi_struct *n) 504 { 505 BUG_ON(!test_bit(NAPI_STATE_SCHED, &n->state)); 506 smp_mb__before_atomic(); 507 clear_bit(NAPI_STATE_SCHED, &n->state); 508 clear_bit(NAPI_STATE_NPSVC, &n->state); 509 } 510 511 /** 512 * napi_synchronize - wait until NAPI is not running 513 * @n: NAPI context 514 * 515 * Wait until NAPI is done being scheduled on this context. 516 * Waits till any outstanding processing completes but 517 * does not disable future activations. 518 */ 519 static inline void napi_synchronize(const struct napi_struct *n) 520 { 521 if (IS_ENABLED(CONFIG_SMP)) 522 while (test_bit(NAPI_STATE_SCHED, &n->state)) 523 msleep(1); 524 else 525 barrier(); 526 } 527 528 enum netdev_queue_state_t { 529 __QUEUE_STATE_DRV_XOFF, 530 __QUEUE_STATE_STACK_XOFF, 531 __QUEUE_STATE_FROZEN, 532 }; 533 534 #define QUEUE_STATE_DRV_XOFF (1 << __QUEUE_STATE_DRV_XOFF) 535 #define QUEUE_STATE_STACK_XOFF (1 << __QUEUE_STATE_STACK_XOFF) 536 #define QUEUE_STATE_FROZEN (1 << __QUEUE_STATE_FROZEN) 537 538 #define QUEUE_STATE_ANY_XOFF (QUEUE_STATE_DRV_XOFF | QUEUE_STATE_STACK_XOFF) 539 #define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \ 540 QUEUE_STATE_FROZEN) 541 #define QUEUE_STATE_DRV_XOFF_OR_FROZEN (QUEUE_STATE_DRV_XOFF | \ 542 QUEUE_STATE_FROZEN) 543 544 /* 545 * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue. The 546 * netif_tx_* functions below are used to manipulate this flag. The 547 * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit 548 * queue independently. The netif_xmit_*stopped functions below are called 549 * to check if the queue has been stopped by the driver or stack (either 550 * of the XOFF bits are set in the state). Drivers should not need to call 551 * netif_xmit*stopped functions, they should only be using netif_tx_*. 552 */ 553 554 struct netdev_queue { 555 /* 556 * read-mostly part 557 */ 558 struct net_device *dev; 559 struct Qdisc __rcu *qdisc; 560 struct Qdisc *qdisc_sleeping; 561 #ifdef CONFIG_SYSFS 562 struct kobject kobj; 563 #endif 564 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA) 565 int numa_node; 566 #endif 567 unsigned long tx_maxrate; 568 /* 569 * Number of TX timeouts for this queue 570 * (/sys/class/net/DEV/Q/trans_timeout) 571 */ 572 unsigned long trans_timeout; 573 /* 574 * write-mostly part 575 */ 576 spinlock_t _xmit_lock ____cacheline_aligned_in_smp; 577 int xmit_lock_owner; 578 /* 579 * Time (in jiffies) of last Tx 580 */ 581 unsigned long trans_start; 582 583 unsigned long state; 584 585 #ifdef CONFIG_BQL 586 struct dql dql; 587 #endif 588 } ____cacheline_aligned_in_smp; 589 590 extern int sysctl_fb_tunnels_only_for_init_net; 591 592 static inline bool net_has_fallback_tunnels(const struct net *net) 593 { 594 return net == &init_net || 595 !IS_ENABLED(CONFIG_SYSCTL) || 596 !sysctl_fb_tunnels_only_for_init_net; 597 } 598 599 static inline int netdev_queue_numa_node_read(const struct netdev_queue *q) 600 { 601 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA) 602 return q->numa_node; 603 #else 604 return NUMA_NO_NODE; 605 #endif 606 } 607 608 static inline void netdev_queue_numa_node_write(struct netdev_queue *q, int node) 609 { 610 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA) 611 q->numa_node = node; 612 #endif 613 } 614 615 #ifdef CONFIG_RPS 616 /* 617 * This structure holds an RPS map which can be of variable length. The 618 * map is an array of CPUs. 619 */ 620 struct rps_map { 621 unsigned int len; 622 struct rcu_head rcu; 623 u16 cpus[0]; 624 }; 625 #define RPS_MAP_SIZE(_num) (sizeof(struct rps_map) + ((_num) * sizeof(u16))) 626 627 /* 628 * The rps_dev_flow structure contains the mapping of a flow to a CPU, the 629 * tail pointer for that CPU's input queue at the time of last enqueue, and 630 * a hardware filter index. 631 */ 632 struct rps_dev_flow { 633 u16 cpu; 634 u16 filter; 635 unsigned int last_qtail; 636 }; 637 #define RPS_NO_FILTER 0xffff 638 639 /* 640 * The rps_dev_flow_table structure contains a table of flow mappings. 641 */ 642 struct rps_dev_flow_table { 643 unsigned int mask; 644 struct rcu_head rcu; 645 struct rps_dev_flow flows[0]; 646 }; 647 #define RPS_DEV_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_dev_flow_table) + \ 648 ((_num) * sizeof(struct rps_dev_flow))) 649 650 /* 651 * The rps_sock_flow_table contains mappings of flows to the last CPU 652 * on which they were processed by the application (set in recvmsg). 653 * Each entry is a 32bit value. Upper part is the high-order bits 654 * of flow hash, lower part is CPU number. 655 * rps_cpu_mask is used to partition the space, depending on number of 656 * possible CPUs : rps_cpu_mask = roundup_pow_of_two(nr_cpu_ids) - 1 657 * For example, if 64 CPUs are possible, rps_cpu_mask = 0x3f, 658 * meaning we use 32-6=26 bits for the hash. 659 */ 660 struct rps_sock_flow_table { 661 u32 mask; 662 663 u32 ents[0] ____cacheline_aligned_in_smp; 664 }; 665 #define RPS_SOCK_FLOW_TABLE_SIZE(_num) (offsetof(struct rps_sock_flow_table, ents[_num])) 666 667 #define RPS_NO_CPU 0xffff 668 669 extern u32 rps_cpu_mask; 670 extern struct rps_sock_flow_table __rcu *rps_sock_flow_table; 671 672 static inline void rps_record_sock_flow(struct rps_sock_flow_table *table, 673 u32 hash) 674 { 675 if (table && hash) { 676 unsigned int index = hash & table->mask; 677 u32 val = hash & ~rps_cpu_mask; 678 679 /* We only give a hint, preemption can change CPU under us */ 680 val |= raw_smp_processor_id(); 681 682 if (table->ents[index] != val) 683 table->ents[index] = val; 684 } 685 } 686 687 #ifdef CONFIG_RFS_ACCEL 688 bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index, u32 flow_id, 689 u16 filter_id); 690 #endif 691 #endif /* CONFIG_RPS */ 692 693 /* This structure contains an instance of an RX queue. */ 694 struct netdev_rx_queue { 695 #ifdef CONFIG_RPS 696 struct rps_map __rcu *rps_map; 697 struct rps_dev_flow_table __rcu *rps_flow_table; 698 #endif 699 struct kobject kobj; 700 struct net_device *dev; 701 struct xdp_rxq_info xdp_rxq; 702 } ____cacheline_aligned_in_smp; 703 704 /* 705 * RX queue sysfs structures and functions. 706 */ 707 struct rx_queue_attribute { 708 struct attribute attr; 709 ssize_t (*show)(struct netdev_rx_queue *queue, char *buf); 710 ssize_t (*store)(struct netdev_rx_queue *queue, 711 const char *buf, size_t len); 712 }; 713 714 #ifdef CONFIG_XPS 715 /* 716 * This structure holds an XPS map which can be of variable length. The 717 * map is an array of queues. 718 */ 719 struct xps_map { 720 unsigned int len; 721 unsigned int alloc_len; 722 struct rcu_head rcu; 723 u16 queues[0]; 724 }; 725 #define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16))) 726 #define XPS_MIN_MAP_ALLOC ((L1_CACHE_ALIGN(offsetof(struct xps_map, queues[1])) \ 727 - sizeof(struct xps_map)) / sizeof(u16)) 728 729 /* 730 * This structure holds all XPS maps for device. Maps are indexed by CPU. 731 */ 732 struct xps_dev_maps { 733 struct rcu_head rcu; 734 struct xps_map __rcu *cpu_map[0]; 735 }; 736 #define XPS_DEV_MAPS_SIZE(_tcs) (sizeof(struct xps_dev_maps) + \ 737 (nr_cpu_ids * (_tcs) * sizeof(struct xps_map *))) 738 #endif /* CONFIG_XPS */ 739 740 #define TC_MAX_QUEUE 16 741 #define TC_BITMASK 15 742 /* HW offloaded queuing disciplines txq count and offset maps */ 743 struct netdev_tc_txq { 744 u16 count; 745 u16 offset; 746 }; 747 748 #if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE) 749 /* 750 * This structure is to hold information about the device 751 * configured to run FCoE protocol stack. 752 */ 753 struct netdev_fcoe_hbainfo { 754 char manufacturer[64]; 755 char serial_number[64]; 756 char hardware_version[64]; 757 char driver_version[64]; 758 char optionrom_version[64]; 759 char firmware_version[64]; 760 char model[256]; 761 char model_description[256]; 762 }; 763 #endif 764 765 #define MAX_PHYS_ITEM_ID_LEN 32 766 767 /* This structure holds a unique identifier to identify some 768 * physical item (port for example) used by a netdevice. 769 */ 770 struct netdev_phys_item_id { 771 unsigned char id[MAX_PHYS_ITEM_ID_LEN]; 772 unsigned char id_len; 773 }; 774 775 static inline bool netdev_phys_item_id_same(struct netdev_phys_item_id *a, 776 struct netdev_phys_item_id *b) 777 { 778 return a->id_len == b->id_len && 779 memcmp(a->id, b->id, a->id_len) == 0; 780 } 781 782 typedef u16 (*select_queue_fallback_t)(struct net_device *dev, 783 struct sk_buff *skb); 784 785 enum tc_setup_type { 786 TC_SETUP_QDISC_MQPRIO, 787 TC_SETUP_CLSU32, 788 TC_SETUP_CLSFLOWER, 789 TC_SETUP_CLSMATCHALL, 790 TC_SETUP_CLSBPF, 791 TC_SETUP_BLOCK, 792 TC_SETUP_QDISC_CBS, 793 TC_SETUP_QDISC_RED, 794 TC_SETUP_QDISC_PRIO, 795 TC_SETUP_QDISC_MQ, 796 }; 797 798 /* These structures hold the attributes of bpf state that are being passed 799 * to the netdevice through the bpf op. 800 */ 801 enum bpf_netdev_command { 802 /* Set or clear a bpf program used in the earliest stages of packet 803 * rx. The prog will have been loaded as BPF_PROG_TYPE_XDP. The callee 804 * is responsible for calling bpf_prog_put on any old progs that are 805 * stored. In case of error, the callee need not release the new prog 806 * reference, but on success it takes ownership and must bpf_prog_put 807 * when it is no longer used. 808 */ 809 XDP_SETUP_PROG, 810 XDP_SETUP_PROG_HW, 811 /* Check if a bpf program is set on the device. The callee should 812 * set @prog_attached to one of XDP_ATTACHED_* values, note that "true" 813 * is equivalent to XDP_ATTACHED_DRV. 814 */ 815 XDP_QUERY_PROG, 816 /* BPF program for offload callbacks, invoked at program load time. */ 817 BPF_OFFLOAD_VERIFIER_PREP, 818 BPF_OFFLOAD_TRANSLATE, 819 BPF_OFFLOAD_DESTROY, 820 BPF_OFFLOAD_MAP_ALLOC, 821 BPF_OFFLOAD_MAP_FREE, 822 XDP_QUERY_XSK_UMEM, 823 XDP_SETUP_XSK_UMEM, 824 }; 825 826 struct bpf_prog_offload_ops; 827 struct netlink_ext_ack; 828 struct xdp_umem; 829 830 struct netdev_bpf { 831 enum bpf_netdev_command command; 832 union { 833 /* XDP_SETUP_PROG */ 834 struct { 835 u32 flags; 836 struct bpf_prog *prog; 837 struct netlink_ext_ack *extack; 838 }; 839 /* XDP_QUERY_PROG */ 840 struct { 841 u8 prog_attached; 842 u32 prog_id; 843 /* flags with which program was installed */ 844 u32 prog_flags; 845 }; 846 /* BPF_OFFLOAD_VERIFIER_PREP */ 847 struct { 848 struct bpf_prog *prog; 849 const struct bpf_prog_offload_ops *ops; /* callee set */ 850 } verifier; 851 /* BPF_OFFLOAD_TRANSLATE, BPF_OFFLOAD_DESTROY */ 852 struct { 853 struct bpf_prog *prog; 854 } offload; 855 /* BPF_OFFLOAD_MAP_ALLOC, BPF_OFFLOAD_MAP_FREE */ 856 struct { 857 struct bpf_offloaded_map *offmap; 858 }; 859 /* XDP_SETUP_XSK_UMEM */ 860 struct { 861 struct xdp_umem *umem; 862 u16 queue_id; 863 } xsk; 864 }; 865 }; 866 867 #ifdef CONFIG_XFRM_OFFLOAD 868 struct xfrmdev_ops { 869 int (*xdo_dev_state_add) (struct xfrm_state *x); 870 void (*xdo_dev_state_delete) (struct xfrm_state *x); 871 void (*xdo_dev_state_free) (struct xfrm_state *x); 872 bool (*xdo_dev_offload_ok) (struct sk_buff *skb, 873 struct xfrm_state *x); 874 void (*xdo_dev_state_advance_esn) (struct xfrm_state *x); 875 }; 876 #endif 877 878 #if IS_ENABLED(CONFIG_TLS_DEVICE) 879 enum tls_offload_ctx_dir { 880 TLS_OFFLOAD_CTX_DIR_RX, 881 TLS_OFFLOAD_CTX_DIR_TX, 882 }; 883 884 struct tls_crypto_info; 885 struct tls_context; 886 887 struct tlsdev_ops { 888 int (*tls_dev_add)(struct net_device *netdev, struct sock *sk, 889 enum tls_offload_ctx_dir direction, 890 struct tls_crypto_info *crypto_info, 891 u32 start_offload_tcp_sn); 892 void (*tls_dev_del)(struct net_device *netdev, 893 struct tls_context *ctx, 894 enum tls_offload_ctx_dir direction); 895 }; 896 #endif 897 898 struct dev_ifalias { 899 struct rcu_head rcuhead; 900 char ifalias[]; 901 }; 902 903 /* 904 * This structure defines the management hooks for network devices. 905 * The following hooks can be defined; unless noted otherwise, they are 906 * optional and can be filled with a null pointer. 907 * 908 * int (*ndo_init)(struct net_device *dev); 909 * This function is called once when a network device is registered. 910 * The network device can use this for any late stage initialization 911 * or semantic validation. It can fail with an error code which will 912 * be propagated back to register_netdev. 913 * 914 * void (*ndo_uninit)(struct net_device *dev); 915 * This function is called when device is unregistered or when registration 916 * fails. It is not called if init fails. 917 * 918 * int (*ndo_open)(struct net_device *dev); 919 * This function is called when a network device transitions to the up 920 * state. 921 * 922 * int (*ndo_stop)(struct net_device *dev); 923 * This function is called when a network device transitions to the down 924 * state. 925 * 926 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb, 927 * struct net_device *dev); 928 * Called when a packet needs to be transmitted. 929 * Returns NETDEV_TX_OK. Can return NETDEV_TX_BUSY, but you should stop 930 * the queue before that can happen; it's for obsolete devices and weird 931 * corner cases, but the stack really does a non-trivial amount 932 * of useless work if you return NETDEV_TX_BUSY. 933 * Required; cannot be NULL. 934 * 935 * netdev_features_t (*ndo_features_check)(struct sk_buff *skb, 936 * struct net_device *dev 937 * netdev_features_t features); 938 * Called by core transmit path to determine if device is capable of 939 * performing offload operations on a given packet. This is to give 940 * the device an opportunity to implement any restrictions that cannot 941 * be otherwise expressed by feature flags. The check is called with 942 * the set of features that the stack has calculated and it returns 943 * those the driver believes to be appropriate. 944 * 945 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb, 946 * void *accel_priv, select_queue_fallback_t fallback); 947 * Called to decide which queue to use when device supports multiple 948 * transmit queues. 949 * 950 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags); 951 * This function is called to allow device receiver to make 952 * changes to configuration when multicast or promiscuous is enabled. 953 * 954 * void (*ndo_set_rx_mode)(struct net_device *dev); 955 * This function is called device changes address list filtering. 956 * If driver handles unicast address filtering, it should set 957 * IFF_UNICAST_FLT in its priv_flags. 958 * 959 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr); 960 * This function is called when the Media Access Control address 961 * needs to be changed. If this interface is not defined, the 962 * MAC address can not be changed. 963 * 964 * int (*ndo_validate_addr)(struct net_device *dev); 965 * Test if Media Access Control address is valid for the device. 966 * 967 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd); 968 * Called when a user requests an ioctl which can't be handled by 969 * the generic interface code. If not defined ioctls return 970 * not supported error code. 971 * 972 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map); 973 * Used to set network devices bus interface parameters. This interface 974 * is retained for legacy reasons; new devices should use the bus 975 * interface (PCI) for low level management. 976 * 977 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu); 978 * Called when a user wants to change the Maximum Transfer Unit 979 * of a device. 980 * 981 * void (*ndo_tx_timeout)(struct net_device *dev); 982 * Callback used when the transmitter has not made any progress 983 * for dev->watchdog ticks. 984 * 985 * void (*ndo_get_stats64)(struct net_device *dev, 986 * struct rtnl_link_stats64 *storage); 987 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev); 988 * Called when a user wants to get the network device usage 989 * statistics. Drivers must do one of the following: 990 * 1. Define @ndo_get_stats64 to fill in a zero-initialised 991 * rtnl_link_stats64 structure passed by the caller. 992 * 2. Define @ndo_get_stats to update a net_device_stats structure 993 * (which should normally be dev->stats) and return a pointer to 994 * it. The structure may be changed asynchronously only if each 995 * field is written atomically. 996 * 3. Update dev->stats asynchronously and atomically, and define 997 * neither operation. 998 * 999 * bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id) 1000 * Return true if this device supports offload stats of this attr_id. 1001 * 1002 * int (*ndo_get_offload_stats)(int attr_id, const struct net_device *dev, 1003 * void *attr_data) 1004 * Get statistics for offload operations by attr_id. Write it into the 1005 * attr_data pointer. 1006 * 1007 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16 vid); 1008 * If device supports VLAN filtering this function is called when a 1009 * VLAN id is registered. 1010 * 1011 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, __be16 proto, u16 vid); 1012 * If device supports VLAN filtering this function is called when a 1013 * VLAN id is unregistered. 1014 * 1015 * void (*ndo_poll_controller)(struct net_device *dev); 1016 * 1017 * SR-IOV management functions. 1018 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac); 1019 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan, 1020 * u8 qos, __be16 proto); 1021 * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate, 1022 * int max_tx_rate); 1023 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting); 1024 * int (*ndo_set_vf_trust)(struct net_device *dev, int vf, bool setting); 1025 * int (*ndo_get_vf_config)(struct net_device *dev, 1026 * int vf, struct ifla_vf_info *ivf); 1027 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state); 1028 * int (*ndo_set_vf_port)(struct net_device *dev, int vf, 1029 * struct nlattr *port[]); 1030 * 1031 * Enable or disable the VF ability to query its RSS Redirection Table and 1032 * Hash Key. This is needed since on some devices VF share this information 1033 * with PF and querying it may introduce a theoretical security risk. 1034 * int (*ndo_set_vf_rss_query_en)(struct net_device *dev, int vf, bool setting); 1035 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb); 1036 * int (*ndo_setup_tc)(struct net_device *dev, enum tc_setup_type type, 1037 * void *type_data); 1038 * Called to setup any 'tc' scheduler, classifier or action on @dev. 1039 * This is always called from the stack with the rtnl lock held and netif 1040 * tx queues stopped. This allows the netdevice to perform queue 1041 * management safely. 1042 * 1043 * Fiber Channel over Ethernet (FCoE) offload functions. 1044 * int (*ndo_fcoe_enable)(struct net_device *dev); 1045 * Called when the FCoE protocol stack wants to start using LLD for FCoE 1046 * so the underlying device can perform whatever needed configuration or 1047 * initialization to support acceleration of FCoE traffic. 1048 * 1049 * int (*ndo_fcoe_disable)(struct net_device *dev); 1050 * Called when the FCoE protocol stack wants to stop using LLD for FCoE 1051 * so the underlying device can perform whatever needed clean-ups to 1052 * stop supporting acceleration of FCoE traffic. 1053 * 1054 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid, 1055 * struct scatterlist *sgl, unsigned int sgc); 1056 * Called when the FCoE Initiator wants to initialize an I/O that 1057 * is a possible candidate for Direct Data Placement (DDP). The LLD can 1058 * perform necessary setup and returns 1 to indicate the device is set up 1059 * successfully to perform DDP on this I/O, otherwise this returns 0. 1060 * 1061 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid); 1062 * Called when the FCoE Initiator/Target is done with the DDPed I/O as 1063 * indicated by the FC exchange id 'xid', so the underlying device can 1064 * clean up and reuse resources for later DDP requests. 1065 * 1066 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid, 1067 * struct scatterlist *sgl, unsigned int sgc); 1068 * Called when the FCoE Target wants to initialize an I/O that 1069 * is a possible candidate for Direct Data Placement (DDP). The LLD can 1070 * perform necessary setup and returns 1 to indicate the device is set up 1071 * successfully to perform DDP on this I/O, otherwise this returns 0. 1072 * 1073 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev, 1074 * struct netdev_fcoe_hbainfo *hbainfo); 1075 * Called when the FCoE Protocol stack wants information on the underlying 1076 * device. This information is utilized by the FCoE protocol stack to 1077 * register attributes with Fiber Channel management service as per the 1078 * FC-GS Fabric Device Management Information(FDMI) specification. 1079 * 1080 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type); 1081 * Called when the underlying device wants to override default World Wide 1082 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own 1083 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE 1084 * protocol stack to use. 1085 * 1086 * RFS acceleration. 1087 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb, 1088 * u16 rxq_index, u32 flow_id); 1089 * Set hardware filter for RFS. rxq_index is the target queue index; 1090 * flow_id is a flow ID to be passed to rps_may_expire_flow() later. 1091 * Return the filter ID on success, or a negative error code. 1092 * 1093 * Slave management functions (for bridge, bonding, etc). 1094 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev); 1095 * Called to make another netdev an underling. 1096 * 1097 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev); 1098 * Called to release previously enslaved netdev. 1099 * 1100 * Feature/offload setting functions. 1101 * netdev_features_t (*ndo_fix_features)(struct net_device *dev, 1102 * netdev_features_t features); 1103 * Adjusts the requested feature flags according to device-specific 1104 * constraints, and returns the resulting flags. Must not modify 1105 * the device state. 1106 * 1107 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features); 1108 * Called to update device configuration to new features. Passed 1109 * feature set might be less than what was returned by ndo_fix_features()). 1110 * Must return >0 or -errno if it changed dev->features itself. 1111 * 1112 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[], 1113 * struct net_device *dev, 1114 * const unsigned char *addr, u16 vid, u16 flags) 1115 * Adds an FDB entry to dev for addr. 1116 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[], 1117 * struct net_device *dev, 1118 * const unsigned char *addr, u16 vid) 1119 * Deletes the FDB entry from dev coresponding to addr. 1120 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb, 1121 * struct net_device *dev, struct net_device *filter_dev, 1122 * int *idx) 1123 * Used to add FDB entries to dump requests. Implementers should add 1124 * entries to skb and update idx with the number of entries. 1125 * 1126 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh, 1127 * u16 flags) 1128 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq, 1129 * struct net_device *dev, u32 filter_mask, 1130 * int nlflags) 1131 * int (*ndo_bridge_dellink)(struct net_device *dev, struct nlmsghdr *nlh, 1132 * u16 flags); 1133 * 1134 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier); 1135 * Called to change device carrier. Soft-devices (like dummy, team, etc) 1136 * which do not represent real hardware may define this to allow their 1137 * userspace components to manage their virtual carrier state. Devices 1138 * that determine carrier state from physical hardware properties (eg 1139 * network cables) or protocol-dependent mechanisms (eg 1140 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function. 1141 * 1142 * int (*ndo_get_phys_port_id)(struct net_device *dev, 1143 * struct netdev_phys_item_id *ppid); 1144 * Called to get ID of physical port of this device. If driver does 1145 * not implement this, it is assumed that the hw is not able to have 1146 * multiple net devices on single physical port. 1147 * 1148 * void (*ndo_udp_tunnel_add)(struct net_device *dev, 1149 * struct udp_tunnel_info *ti); 1150 * Called by UDP tunnel to notify a driver about the UDP port and socket 1151 * address family that a UDP tunnel is listnening to. It is called only 1152 * when a new port starts listening. The operation is protected by the 1153 * RTNL. 1154 * 1155 * void (*ndo_udp_tunnel_del)(struct net_device *dev, 1156 * struct udp_tunnel_info *ti); 1157 * Called by UDP tunnel to notify the driver about a UDP port and socket 1158 * address family that the UDP tunnel is not listening to anymore. The 1159 * operation is protected by the RTNL. 1160 * 1161 * void* (*ndo_dfwd_add_station)(struct net_device *pdev, 1162 * struct net_device *dev) 1163 * Called by upper layer devices to accelerate switching or other 1164 * station functionality into hardware. 'pdev is the lowerdev 1165 * to use for the offload and 'dev' is the net device that will 1166 * back the offload. Returns a pointer to the private structure 1167 * the upper layer will maintain. 1168 * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv) 1169 * Called by upper layer device to delete the station created 1170 * by 'ndo_dfwd_add_station'. 'pdev' is the net device backing 1171 * the station and priv is the structure returned by the add 1172 * operation. 1173 * int (*ndo_set_tx_maxrate)(struct net_device *dev, 1174 * int queue_index, u32 maxrate); 1175 * Called when a user wants to set a max-rate limitation of specific 1176 * TX queue. 1177 * int (*ndo_get_iflink)(const struct net_device *dev); 1178 * Called to get the iflink value of this device. 1179 * void (*ndo_change_proto_down)(struct net_device *dev, 1180 * bool proto_down); 1181 * This function is used to pass protocol port error state information 1182 * to the switch driver. The switch driver can react to the proto_down 1183 * by doing a phys down on the associated switch port. 1184 * int (*ndo_fill_metadata_dst)(struct net_device *dev, struct sk_buff *skb); 1185 * This function is used to get egress tunnel information for given skb. 1186 * This is useful for retrieving outer tunnel header parameters while 1187 * sampling packet. 1188 * void (*ndo_set_rx_headroom)(struct net_device *dev, int needed_headroom); 1189 * This function is used to specify the headroom that the skb must 1190 * consider when allocation skb during packet reception. Setting 1191 * appropriate rx headroom value allows avoiding skb head copy on 1192 * forward. Setting a negative value resets the rx headroom to the 1193 * default value. 1194 * int (*ndo_bpf)(struct net_device *dev, struct netdev_bpf *bpf); 1195 * This function is used to set or query state related to XDP on the 1196 * netdevice and manage BPF offload. See definition of 1197 * enum bpf_netdev_command for details. 1198 * int (*ndo_xdp_xmit)(struct net_device *dev, int n, struct xdp_frame **xdp, 1199 * u32 flags); 1200 * This function is used to submit @n XDP packets for transmit on a 1201 * netdevice. Returns number of frames successfully transmitted, frames 1202 * that got dropped are freed/returned via xdp_return_frame(). 1203 * Returns negative number, means general error invoking ndo, meaning 1204 * no frames were xmit'ed and core-caller will free all frames. 1205 */ 1206 struct net_device_ops { 1207 int (*ndo_init)(struct net_device *dev); 1208 void (*ndo_uninit)(struct net_device *dev); 1209 int (*ndo_open)(struct net_device *dev); 1210 int (*ndo_stop)(struct net_device *dev); 1211 netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb, 1212 struct net_device *dev); 1213 netdev_features_t (*ndo_features_check)(struct sk_buff *skb, 1214 struct net_device *dev, 1215 netdev_features_t features); 1216 u16 (*ndo_select_queue)(struct net_device *dev, 1217 struct sk_buff *skb, 1218 void *accel_priv, 1219 select_queue_fallback_t fallback); 1220 void (*ndo_change_rx_flags)(struct net_device *dev, 1221 int flags); 1222 void (*ndo_set_rx_mode)(struct net_device *dev); 1223 int (*ndo_set_mac_address)(struct net_device *dev, 1224 void *addr); 1225 int (*ndo_validate_addr)(struct net_device *dev); 1226 int (*ndo_do_ioctl)(struct net_device *dev, 1227 struct ifreq *ifr, int cmd); 1228 int (*ndo_set_config)(struct net_device *dev, 1229 struct ifmap *map); 1230 int (*ndo_change_mtu)(struct net_device *dev, 1231 int new_mtu); 1232 int (*ndo_neigh_setup)(struct net_device *dev, 1233 struct neigh_parms *); 1234 void (*ndo_tx_timeout) (struct net_device *dev); 1235 1236 void (*ndo_get_stats64)(struct net_device *dev, 1237 struct rtnl_link_stats64 *storage); 1238 bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id); 1239 int (*ndo_get_offload_stats)(int attr_id, 1240 const struct net_device *dev, 1241 void *attr_data); 1242 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev); 1243 1244 int (*ndo_vlan_rx_add_vid)(struct net_device *dev, 1245 __be16 proto, u16 vid); 1246 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, 1247 __be16 proto, u16 vid); 1248 #ifdef CONFIG_NET_POLL_CONTROLLER 1249 void (*ndo_poll_controller)(struct net_device *dev); 1250 int (*ndo_netpoll_setup)(struct net_device *dev, 1251 struct netpoll_info *info); 1252 void (*ndo_netpoll_cleanup)(struct net_device *dev); 1253 #endif 1254 int (*ndo_set_vf_mac)(struct net_device *dev, 1255 int queue, u8 *mac); 1256 int (*ndo_set_vf_vlan)(struct net_device *dev, 1257 int queue, u16 vlan, 1258 u8 qos, __be16 proto); 1259 int (*ndo_set_vf_rate)(struct net_device *dev, 1260 int vf, int min_tx_rate, 1261 int max_tx_rate); 1262 int (*ndo_set_vf_spoofchk)(struct net_device *dev, 1263 int vf, bool setting); 1264 int (*ndo_set_vf_trust)(struct net_device *dev, 1265 int vf, bool setting); 1266 int (*ndo_get_vf_config)(struct net_device *dev, 1267 int vf, 1268 struct ifla_vf_info *ivf); 1269 int (*ndo_set_vf_link_state)(struct net_device *dev, 1270 int vf, int link_state); 1271 int (*ndo_get_vf_stats)(struct net_device *dev, 1272 int vf, 1273 struct ifla_vf_stats 1274 *vf_stats); 1275 int (*ndo_set_vf_port)(struct net_device *dev, 1276 int vf, 1277 struct nlattr *port[]); 1278 int (*ndo_get_vf_port)(struct net_device *dev, 1279 int vf, struct sk_buff *skb); 1280 int (*ndo_set_vf_guid)(struct net_device *dev, 1281 int vf, u64 guid, 1282 int guid_type); 1283 int (*ndo_set_vf_rss_query_en)( 1284 struct net_device *dev, 1285 int vf, bool setting); 1286 int (*ndo_setup_tc)(struct net_device *dev, 1287 enum tc_setup_type type, 1288 void *type_data); 1289 #if IS_ENABLED(CONFIG_FCOE) 1290 int (*ndo_fcoe_enable)(struct net_device *dev); 1291 int (*ndo_fcoe_disable)(struct net_device *dev); 1292 int (*ndo_fcoe_ddp_setup)(struct net_device *dev, 1293 u16 xid, 1294 struct scatterlist *sgl, 1295 unsigned int sgc); 1296 int (*ndo_fcoe_ddp_done)(struct net_device *dev, 1297 u16 xid); 1298 int (*ndo_fcoe_ddp_target)(struct net_device *dev, 1299 u16 xid, 1300 struct scatterlist *sgl, 1301 unsigned int sgc); 1302 int (*ndo_fcoe_get_hbainfo)(struct net_device *dev, 1303 struct netdev_fcoe_hbainfo *hbainfo); 1304 #endif 1305 1306 #if IS_ENABLED(CONFIG_LIBFCOE) 1307 #define NETDEV_FCOE_WWNN 0 1308 #define NETDEV_FCOE_WWPN 1 1309 int (*ndo_fcoe_get_wwn)(struct net_device *dev, 1310 u64 *wwn, int type); 1311 #endif 1312 1313 #ifdef CONFIG_RFS_ACCEL 1314 int (*ndo_rx_flow_steer)(struct net_device *dev, 1315 const struct sk_buff *skb, 1316 u16 rxq_index, 1317 u32 flow_id); 1318 #endif 1319 int (*ndo_add_slave)(struct net_device *dev, 1320 struct net_device *slave_dev, 1321 struct netlink_ext_ack *extack); 1322 int (*ndo_del_slave)(struct net_device *dev, 1323 struct net_device *slave_dev); 1324 netdev_features_t (*ndo_fix_features)(struct net_device *dev, 1325 netdev_features_t features); 1326 int (*ndo_set_features)(struct net_device *dev, 1327 netdev_features_t features); 1328 int (*ndo_neigh_construct)(struct net_device *dev, 1329 struct neighbour *n); 1330 void (*ndo_neigh_destroy)(struct net_device *dev, 1331 struct neighbour *n); 1332 1333 int (*ndo_fdb_add)(struct ndmsg *ndm, 1334 struct nlattr *tb[], 1335 struct net_device *dev, 1336 const unsigned char *addr, 1337 u16 vid, 1338 u16 flags); 1339 int (*ndo_fdb_del)(struct ndmsg *ndm, 1340 struct nlattr *tb[], 1341 struct net_device *dev, 1342 const unsigned char *addr, 1343 u16 vid); 1344 int (*ndo_fdb_dump)(struct sk_buff *skb, 1345 struct netlink_callback *cb, 1346 struct net_device *dev, 1347 struct net_device *filter_dev, 1348 int *idx); 1349 1350 int (*ndo_bridge_setlink)(struct net_device *dev, 1351 struct nlmsghdr *nlh, 1352 u16 flags); 1353 int (*ndo_bridge_getlink)(struct sk_buff *skb, 1354 u32 pid, u32 seq, 1355 struct net_device *dev, 1356 u32 filter_mask, 1357 int nlflags); 1358 int (*ndo_bridge_dellink)(struct net_device *dev, 1359 struct nlmsghdr *nlh, 1360 u16 flags); 1361 int (*ndo_change_carrier)(struct net_device *dev, 1362 bool new_carrier); 1363 int (*ndo_get_phys_port_id)(struct net_device *dev, 1364 struct netdev_phys_item_id *ppid); 1365 int (*ndo_get_phys_port_name)(struct net_device *dev, 1366 char *name, size_t len); 1367 void (*ndo_udp_tunnel_add)(struct net_device *dev, 1368 struct udp_tunnel_info *ti); 1369 void (*ndo_udp_tunnel_del)(struct net_device *dev, 1370 struct udp_tunnel_info *ti); 1371 void* (*ndo_dfwd_add_station)(struct net_device *pdev, 1372 struct net_device *dev); 1373 void (*ndo_dfwd_del_station)(struct net_device *pdev, 1374 void *priv); 1375 1376 int (*ndo_get_lock_subclass)(struct net_device *dev); 1377 int (*ndo_set_tx_maxrate)(struct net_device *dev, 1378 int queue_index, 1379 u32 maxrate); 1380 int (*ndo_get_iflink)(const struct net_device *dev); 1381 int (*ndo_change_proto_down)(struct net_device *dev, 1382 bool proto_down); 1383 int (*ndo_fill_metadata_dst)(struct net_device *dev, 1384 struct sk_buff *skb); 1385 void (*ndo_set_rx_headroom)(struct net_device *dev, 1386 int needed_headroom); 1387 int (*ndo_bpf)(struct net_device *dev, 1388 struct netdev_bpf *bpf); 1389 int (*ndo_xdp_xmit)(struct net_device *dev, int n, 1390 struct xdp_frame **xdp, 1391 u32 flags); 1392 int (*ndo_xsk_async_xmit)(struct net_device *dev, 1393 u32 queue_id); 1394 }; 1395 1396 /** 1397 * enum net_device_priv_flags - &struct net_device priv_flags 1398 * 1399 * These are the &struct net_device, they are only set internally 1400 * by drivers and used in the kernel. These flags are invisible to 1401 * userspace; this means that the order of these flags can change 1402 * during any kernel release. 1403 * 1404 * You should have a pretty good reason to be extending these flags. 1405 * 1406 * @IFF_802_1Q_VLAN: 802.1Q VLAN device 1407 * @IFF_EBRIDGE: Ethernet bridging device 1408 * @IFF_BONDING: bonding master or slave 1409 * @IFF_ISATAP: ISATAP interface (RFC4214) 1410 * @IFF_WAN_HDLC: WAN HDLC device 1411 * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to 1412 * release skb->dst 1413 * @IFF_DONT_BRIDGE: disallow bridging this ether dev 1414 * @IFF_DISABLE_NETPOLL: disable netpoll at run-time 1415 * @IFF_MACVLAN_PORT: device used as macvlan port 1416 * @IFF_BRIDGE_PORT: device used as bridge port 1417 * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port 1418 * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit 1419 * @IFF_UNICAST_FLT: Supports unicast filtering 1420 * @IFF_TEAM_PORT: device used as team port 1421 * @IFF_SUPP_NOFCS: device supports sending custom FCS 1422 * @IFF_LIVE_ADDR_CHANGE: device supports hardware address 1423 * change when it's running 1424 * @IFF_MACVLAN: Macvlan device 1425 * @IFF_XMIT_DST_RELEASE_PERM: IFF_XMIT_DST_RELEASE not taking into account 1426 * underlying stacked devices 1427 * @IFF_L3MDEV_MASTER: device is an L3 master device 1428 * @IFF_NO_QUEUE: device can run without qdisc attached 1429 * @IFF_OPENVSWITCH: device is a Open vSwitch master 1430 * @IFF_L3MDEV_SLAVE: device is enslaved to an L3 master device 1431 * @IFF_TEAM: device is a team device 1432 * @IFF_RXFH_CONFIGURED: device has had Rx Flow indirection table configured 1433 * @IFF_PHONY_HEADROOM: the headroom value is controlled by an external 1434 * entity (i.e. the master device for bridged veth) 1435 * @IFF_MACSEC: device is a MACsec device 1436 * @IFF_NO_RX_HANDLER: device doesn't support the rx_handler hook 1437 * @IFF_FAILOVER: device is a failover master device 1438 * @IFF_FAILOVER_SLAVE: device is lower dev of a failover master device 1439 */ 1440 enum netdev_priv_flags { 1441 IFF_802_1Q_VLAN = 1<<0, 1442 IFF_EBRIDGE = 1<<1, 1443 IFF_BONDING = 1<<2, 1444 IFF_ISATAP = 1<<3, 1445 IFF_WAN_HDLC = 1<<4, 1446 IFF_XMIT_DST_RELEASE = 1<<5, 1447 IFF_DONT_BRIDGE = 1<<6, 1448 IFF_DISABLE_NETPOLL = 1<<7, 1449 IFF_MACVLAN_PORT = 1<<8, 1450 IFF_BRIDGE_PORT = 1<<9, 1451 IFF_OVS_DATAPATH = 1<<10, 1452 IFF_TX_SKB_SHARING = 1<<11, 1453 IFF_UNICAST_FLT = 1<<12, 1454 IFF_TEAM_PORT = 1<<13, 1455 IFF_SUPP_NOFCS = 1<<14, 1456 IFF_LIVE_ADDR_CHANGE = 1<<15, 1457 IFF_MACVLAN = 1<<16, 1458 IFF_XMIT_DST_RELEASE_PERM = 1<<17, 1459 IFF_L3MDEV_MASTER = 1<<18, 1460 IFF_NO_QUEUE = 1<<19, 1461 IFF_OPENVSWITCH = 1<<20, 1462 IFF_L3MDEV_SLAVE = 1<<21, 1463 IFF_TEAM = 1<<22, 1464 IFF_RXFH_CONFIGURED = 1<<23, 1465 IFF_PHONY_HEADROOM = 1<<24, 1466 IFF_MACSEC = 1<<25, 1467 IFF_NO_RX_HANDLER = 1<<26, 1468 IFF_FAILOVER = 1<<27, 1469 IFF_FAILOVER_SLAVE = 1<<28, 1470 }; 1471 1472 #define IFF_802_1Q_VLAN IFF_802_1Q_VLAN 1473 #define IFF_EBRIDGE IFF_EBRIDGE 1474 #define IFF_BONDING IFF_BONDING 1475 #define IFF_ISATAP IFF_ISATAP 1476 #define IFF_WAN_HDLC IFF_WAN_HDLC 1477 #define IFF_XMIT_DST_RELEASE IFF_XMIT_DST_RELEASE 1478 #define IFF_DONT_BRIDGE IFF_DONT_BRIDGE 1479 #define IFF_DISABLE_NETPOLL IFF_DISABLE_NETPOLL 1480 #define IFF_MACVLAN_PORT IFF_MACVLAN_PORT 1481 #define IFF_BRIDGE_PORT IFF_BRIDGE_PORT 1482 #define IFF_OVS_DATAPATH IFF_OVS_DATAPATH 1483 #define IFF_TX_SKB_SHARING IFF_TX_SKB_SHARING 1484 #define IFF_UNICAST_FLT IFF_UNICAST_FLT 1485 #define IFF_TEAM_PORT IFF_TEAM_PORT 1486 #define IFF_SUPP_NOFCS IFF_SUPP_NOFCS 1487 #define IFF_LIVE_ADDR_CHANGE IFF_LIVE_ADDR_CHANGE 1488 #define IFF_MACVLAN IFF_MACVLAN 1489 #define IFF_XMIT_DST_RELEASE_PERM IFF_XMIT_DST_RELEASE_PERM 1490 #define IFF_L3MDEV_MASTER IFF_L3MDEV_MASTER 1491 #define IFF_NO_QUEUE IFF_NO_QUEUE 1492 #define IFF_OPENVSWITCH IFF_OPENVSWITCH 1493 #define IFF_L3MDEV_SLAVE IFF_L3MDEV_SLAVE 1494 #define IFF_TEAM IFF_TEAM 1495 #define IFF_RXFH_CONFIGURED IFF_RXFH_CONFIGURED 1496 #define IFF_MACSEC IFF_MACSEC 1497 #define IFF_NO_RX_HANDLER IFF_NO_RX_HANDLER 1498 #define IFF_FAILOVER IFF_FAILOVER 1499 #define IFF_FAILOVER_SLAVE IFF_FAILOVER_SLAVE 1500 1501 /** 1502 * struct net_device - The DEVICE structure. 1503 * 1504 * Actually, this whole structure is a big mistake. It mixes I/O 1505 * data with strictly "high-level" data, and it has to know about 1506 * almost every data structure used in the INET module. 1507 * 1508 * @name: This is the first field of the "visible" part of this structure 1509 * (i.e. as seen by users in the "Space.c" file). It is the name 1510 * of the interface. 1511 * 1512 * @name_hlist: Device name hash chain, please keep it close to name[] 1513 * @ifalias: SNMP alias 1514 * @mem_end: Shared memory end 1515 * @mem_start: Shared memory start 1516 * @base_addr: Device I/O address 1517 * @irq: Device IRQ number 1518 * 1519 * @state: Generic network queuing layer state, see netdev_state_t 1520 * @dev_list: The global list of network devices 1521 * @napi_list: List entry used for polling NAPI devices 1522 * @unreg_list: List entry when we are unregistering the 1523 * device; see the function unregister_netdev 1524 * @close_list: List entry used when we are closing the device 1525 * @ptype_all: Device-specific packet handlers for all protocols 1526 * @ptype_specific: Device-specific, protocol-specific packet handlers 1527 * 1528 * @adj_list: Directly linked devices, like slaves for bonding 1529 * @features: Currently active device features 1530 * @hw_features: User-changeable features 1531 * 1532 * @wanted_features: User-requested features 1533 * @vlan_features: Mask of features inheritable by VLAN devices 1534 * 1535 * @hw_enc_features: Mask of features inherited by encapsulating devices 1536 * This field indicates what encapsulation 1537 * offloads the hardware is capable of doing, 1538 * and drivers will need to set them appropriately. 1539 * 1540 * @mpls_features: Mask of features inheritable by MPLS 1541 * 1542 * @ifindex: interface index 1543 * @group: The group the device belongs to 1544 * 1545 * @stats: Statistics struct, which was left as a legacy, use 1546 * rtnl_link_stats64 instead 1547 * 1548 * @rx_dropped: Dropped packets by core network, 1549 * do not use this in drivers 1550 * @tx_dropped: Dropped packets by core network, 1551 * do not use this in drivers 1552 * @rx_nohandler: nohandler dropped packets by core network on 1553 * inactive devices, do not use this in drivers 1554 * @carrier_up_count: Number of times the carrier has been up 1555 * @carrier_down_count: Number of times the carrier has been down 1556 * 1557 * @wireless_handlers: List of functions to handle Wireless Extensions, 1558 * instead of ioctl, 1559 * see <net/iw_handler.h> for details. 1560 * @wireless_data: Instance data managed by the core of wireless extensions 1561 * 1562 * @netdev_ops: Includes several pointers to callbacks, 1563 * if one wants to override the ndo_*() functions 1564 * @ethtool_ops: Management operations 1565 * @ndisc_ops: Includes callbacks for different IPv6 neighbour 1566 * discovery handling. Necessary for e.g. 6LoWPAN. 1567 * @header_ops: Includes callbacks for creating,parsing,caching,etc 1568 * of Layer 2 headers. 1569 * 1570 * @flags: Interface flags (a la BSD) 1571 * @priv_flags: Like 'flags' but invisible to userspace, 1572 * see if.h for the definitions 1573 * @gflags: Global flags ( kept as legacy ) 1574 * @padded: How much padding added by alloc_netdev() 1575 * @operstate: RFC2863 operstate 1576 * @link_mode: Mapping policy to operstate 1577 * @if_port: Selectable AUI, TP, ... 1578 * @dma: DMA channel 1579 * @mtu: Interface MTU value 1580 * @min_mtu: Interface Minimum MTU value 1581 * @max_mtu: Interface Maximum MTU value 1582 * @type: Interface hardware type 1583 * @hard_header_len: Maximum hardware header length. 1584 * @min_header_len: Minimum hardware header length 1585 * 1586 * @needed_headroom: Extra headroom the hardware may need, but not in all 1587 * cases can this be guaranteed 1588 * @needed_tailroom: Extra tailroom the hardware may need, but not in all 1589 * cases can this be guaranteed. Some cases also use 1590 * LL_MAX_HEADER instead to allocate the skb 1591 * 1592 * interface address info: 1593 * 1594 * @perm_addr: Permanent hw address 1595 * @addr_assign_type: Hw address assignment type 1596 * @addr_len: Hardware address length 1597 * @neigh_priv_len: Used in neigh_alloc() 1598 * @dev_id: Used to differentiate devices that share 1599 * the same link layer address 1600 * @dev_port: Used to differentiate devices that share 1601 * the same function 1602 * @addr_list_lock: XXX: need comments on this one 1603 * @uc_promisc: Counter that indicates promiscuous mode 1604 * has been enabled due to the need to listen to 1605 * additional unicast addresses in a device that 1606 * does not implement ndo_set_rx_mode() 1607 * @uc: unicast mac addresses 1608 * @mc: multicast mac addresses 1609 * @dev_addrs: list of device hw addresses 1610 * @queues_kset: Group of all Kobjects in the Tx and RX queues 1611 * @promiscuity: Number of times the NIC is told to work in 1612 * promiscuous mode; if it becomes 0 the NIC will 1613 * exit promiscuous mode 1614 * @allmulti: Counter, enables or disables allmulticast mode 1615 * 1616 * @vlan_info: VLAN info 1617 * @dsa_ptr: dsa specific data 1618 * @tipc_ptr: TIPC specific data 1619 * @atalk_ptr: AppleTalk link 1620 * @ip_ptr: IPv4 specific data 1621 * @dn_ptr: DECnet specific data 1622 * @ip6_ptr: IPv6 specific data 1623 * @ax25_ptr: AX.25 specific data 1624 * @ieee80211_ptr: IEEE 802.11 specific data, assign before registering 1625 * 1626 * @dev_addr: Hw address (before bcast, 1627 * because most packets are unicast) 1628 * 1629 * @_rx: Array of RX queues 1630 * @num_rx_queues: Number of RX queues 1631 * allocated at register_netdev() time 1632 * @real_num_rx_queues: Number of RX queues currently active in device 1633 * 1634 * @rx_handler: handler for received packets 1635 * @rx_handler_data: XXX: need comments on this one 1636 * @miniq_ingress: ingress/clsact qdisc specific data for 1637 * ingress processing 1638 * @ingress_queue: XXX: need comments on this one 1639 * @broadcast: hw bcast address 1640 * 1641 * @rx_cpu_rmap: CPU reverse-mapping for RX completion interrupts, 1642 * indexed by RX queue number. Assigned by driver. 1643 * This must only be set if the ndo_rx_flow_steer 1644 * operation is defined 1645 * @index_hlist: Device index hash chain 1646 * 1647 * @_tx: Array of TX queues 1648 * @num_tx_queues: Number of TX queues allocated at alloc_netdev_mq() time 1649 * @real_num_tx_queues: Number of TX queues currently active in device 1650 * @qdisc: Root qdisc from userspace point of view 1651 * @tx_queue_len: Max frames per queue allowed 1652 * @tx_global_lock: XXX: need comments on this one 1653 * 1654 * @xps_maps: XXX: need comments on this one 1655 * @miniq_egress: clsact qdisc specific data for 1656 * egress processing 1657 * @watchdog_timeo: Represents the timeout that is used by 1658 * the watchdog (see dev_watchdog()) 1659 * @watchdog_timer: List of timers 1660 * 1661 * @pcpu_refcnt: Number of references to this device 1662 * @todo_list: Delayed register/unregister 1663 * @link_watch_list: XXX: need comments on this one 1664 * 1665 * @reg_state: Register/unregister state machine 1666 * @dismantle: Device is going to be freed 1667 * @rtnl_link_state: This enum represents the phases of creating 1668 * a new link 1669 * 1670 * @needs_free_netdev: Should unregister perform free_netdev? 1671 * @priv_destructor: Called from unregister 1672 * @npinfo: XXX: need comments on this one 1673 * @nd_net: Network namespace this network device is inside 1674 * 1675 * @ml_priv: Mid-layer private 1676 * @lstats: Loopback statistics 1677 * @tstats: Tunnel statistics 1678 * @dstats: Dummy statistics 1679 * @vstats: Virtual ethernet statistics 1680 * 1681 * @garp_port: GARP 1682 * @mrp_port: MRP 1683 * 1684 * @dev: Class/net/name entry 1685 * @sysfs_groups: Space for optional device, statistics and wireless 1686 * sysfs groups 1687 * 1688 * @sysfs_rx_queue_group: Space for optional per-rx queue attributes 1689 * @rtnl_link_ops: Rtnl_link_ops 1690 * 1691 * @gso_max_size: Maximum size of generic segmentation offload 1692 * @gso_max_segs: Maximum number of segments that can be passed to the 1693 * NIC for GSO 1694 * 1695 * @dcbnl_ops: Data Center Bridging netlink ops 1696 * @num_tc: Number of traffic classes in the net device 1697 * @tc_to_txq: XXX: need comments on this one 1698 * @prio_tc_map: XXX: need comments on this one 1699 * 1700 * @fcoe_ddp_xid: Max exchange id for FCoE LRO by ddp 1701 * 1702 * @priomap: XXX: need comments on this one 1703 * @phydev: Physical device may attach itself 1704 * for hardware timestamping 1705 * @sfp_bus: attached &struct sfp_bus structure. 1706 * 1707 * @qdisc_tx_busylock: lockdep class annotating Qdisc->busylock spinlock 1708 * @qdisc_running_key: lockdep class annotating Qdisc->running seqcount 1709 * 1710 * @proto_down: protocol port state information can be sent to the 1711 * switch driver and used to set the phys state of the 1712 * switch port. 1713 * 1714 * FIXME: cleanup struct net_device such that network protocol info 1715 * moves out. 1716 */ 1717 1718 struct net_device { 1719 char name[IFNAMSIZ]; 1720 struct hlist_node name_hlist; 1721 struct dev_ifalias __rcu *ifalias; 1722 /* 1723 * I/O specific fields 1724 * FIXME: Merge these and struct ifmap into one 1725 */ 1726 unsigned long mem_end; 1727 unsigned long mem_start; 1728 unsigned long base_addr; 1729 int irq; 1730 1731 /* 1732 * Some hardware also needs these fields (state,dev_list, 1733 * napi_list,unreg_list,close_list) but they are not 1734 * part of the usual set specified in Space.c. 1735 */ 1736 1737 unsigned long state; 1738 1739 struct list_head dev_list; 1740 struct list_head napi_list; 1741 struct list_head unreg_list; 1742 struct list_head close_list; 1743 struct list_head ptype_all; 1744 struct list_head ptype_specific; 1745 1746 struct { 1747 struct list_head upper; 1748 struct list_head lower; 1749 } adj_list; 1750 1751 netdev_features_t features; 1752 netdev_features_t hw_features; 1753 netdev_features_t wanted_features; 1754 netdev_features_t vlan_features; 1755 netdev_features_t hw_enc_features; 1756 netdev_features_t mpls_features; 1757 netdev_features_t gso_partial_features; 1758 1759 int ifindex; 1760 int group; 1761 1762 struct net_device_stats stats; 1763 1764 atomic_long_t rx_dropped; 1765 atomic_long_t tx_dropped; 1766 atomic_long_t rx_nohandler; 1767 1768 /* Stats to monitor link on/off, flapping */ 1769 atomic_t carrier_up_count; 1770 atomic_t carrier_down_count; 1771 1772 #ifdef CONFIG_WIRELESS_EXT 1773 const struct iw_handler_def *wireless_handlers; 1774 struct iw_public_data *wireless_data; 1775 #endif 1776 const struct net_device_ops *netdev_ops; 1777 const struct ethtool_ops *ethtool_ops; 1778 #ifdef CONFIG_NET_SWITCHDEV 1779 const struct switchdev_ops *switchdev_ops; 1780 #endif 1781 #ifdef CONFIG_NET_L3_MASTER_DEV 1782 const struct l3mdev_ops *l3mdev_ops; 1783 #endif 1784 #if IS_ENABLED(CONFIG_IPV6) 1785 const struct ndisc_ops *ndisc_ops; 1786 #endif 1787 1788 #ifdef CONFIG_XFRM_OFFLOAD 1789 const struct xfrmdev_ops *xfrmdev_ops; 1790 #endif 1791 1792 #if IS_ENABLED(CONFIG_TLS_DEVICE) 1793 const struct tlsdev_ops *tlsdev_ops; 1794 #endif 1795 1796 const struct header_ops *header_ops; 1797 1798 unsigned int flags; 1799 unsigned int priv_flags; 1800 1801 unsigned short gflags; 1802 unsigned short padded; 1803 1804 unsigned char operstate; 1805 unsigned char link_mode; 1806 1807 unsigned char if_port; 1808 unsigned char dma; 1809 1810 unsigned int mtu; 1811 unsigned int min_mtu; 1812 unsigned int max_mtu; 1813 unsigned short type; 1814 unsigned short hard_header_len; 1815 unsigned char min_header_len; 1816 1817 unsigned short needed_headroom; 1818 unsigned short needed_tailroom; 1819 1820 /* Interface address info. */ 1821 unsigned char perm_addr[MAX_ADDR_LEN]; 1822 unsigned char addr_assign_type; 1823 unsigned char addr_len; 1824 unsigned short neigh_priv_len; 1825 unsigned short dev_id; 1826 unsigned short dev_port; 1827 spinlock_t addr_list_lock; 1828 unsigned char name_assign_type; 1829 bool uc_promisc; 1830 struct netdev_hw_addr_list uc; 1831 struct netdev_hw_addr_list mc; 1832 struct netdev_hw_addr_list dev_addrs; 1833 1834 #ifdef CONFIG_SYSFS 1835 struct kset *queues_kset; 1836 #endif 1837 unsigned int promiscuity; 1838 unsigned int allmulti; 1839 1840 1841 /* Protocol-specific pointers */ 1842 1843 #if IS_ENABLED(CONFIG_VLAN_8021Q) 1844 struct vlan_info __rcu *vlan_info; 1845 #endif 1846 #if IS_ENABLED(CONFIG_NET_DSA) 1847 struct dsa_port *dsa_ptr; 1848 #endif 1849 #if IS_ENABLED(CONFIG_TIPC) 1850 struct tipc_bearer __rcu *tipc_ptr; 1851 #endif 1852 #if IS_ENABLED(CONFIG_IRDA) || IS_ENABLED(CONFIG_ATALK) 1853 void *atalk_ptr; 1854 #endif 1855 struct in_device __rcu *ip_ptr; 1856 #if IS_ENABLED(CONFIG_DECNET) 1857 struct dn_dev __rcu *dn_ptr; 1858 #endif 1859 struct inet6_dev __rcu *ip6_ptr; 1860 #if IS_ENABLED(CONFIG_AX25) 1861 void *ax25_ptr; 1862 #endif 1863 struct wireless_dev *ieee80211_ptr; 1864 struct wpan_dev *ieee802154_ptr; 1865 #if IS_ENABLED(CONFIG_MPLS_ROUTING) 1866 struct mpls_dev __rcu *mpls_ptr; 1867 #endif 1868 1869 /* 1870 * Cache lines mostly used on receive path (including eth_type_trans()) 1871 */ 1872 /* Interface address info used in eth_type_trans() */ 1873 unsigned char *dev_addr; 1874 1875 struct netdev_rx_queue *_rx; 1876 unsigned int num_rx_queues; 1877 unsigned int real_num_rx_queues; 1878 1879 struct bpf_prog __rcu *xdp_prog; 1880 unsigned long gro_flush_timeout; 1881 rx_handler_func_t __rcu *rx_handler; 1882 void __rcu *rx_handler_data; 1883 1884 #ifdef CONFIG_NET_CLS_ACT 1885 struct mini_Qdisc __rcu *miniq_ingress; 1886 #endif 1887 struct netdev_queue __rcu *ingress_queue; 1888 #ifdef CONFIG_NETFILTER_INGRESS 1889 struct nf_hook_entries __rcu *nf_hooks_ingress; 1890 #endif 1891 1892 unsigned char broadcast[MAX_ADDR_LEN]; 1893 #ifdef CONFIG_RFS_ACCEL 1894 struct cpu_rmap *rx_cpu_rmap; 1895 #endif 1896 struct hlist_node index_hlist; 1897 1898 /* 1899 * Cache lines mostly used on transmit path 1900 */ 1901 struct netdev_queue *_tx ____cacheline_aligned_in_smp; 1902 unsigned int num_tx_queues; 1903 unsigned int real_num_tx_queues; 1904 struct Qdisc *qdisc; 1905 #ifdef CONFIG_NET_SCHED 1906 DECLARE_HASHTABLE (qdisc_hash, 4); 1907 #endif 1908 unsigned int tx_queue_len; 1909 spinlock_t tx_global_lock; 1910 int watchdog_timeo; 1911 1912 #ifdef CONFIG_XPS 1913 struct xps_dev_maps __rcu *xps_maps; 1914 #endif 1915 #ifdef CONFIG_NET_CLS_ACT 1916 struct mini_Qdisc __rcu *miniq_egress; 1917 #endif 1918 1919 /* These may be needed for future network-power-down code. */ 1920 struct timer_list watchdog_timer; 1921 1922 int __percpu *pcpu_refcnt; 1923 struct list_head todo_list; 1924 1925 struct list_head link_watch_list; 1926 1927 enum { NETREG_UNINITIALIZED=0, 1928 NETREG_REGISTERED, /* completed register_netdevice */ 1929 NETREG_UNREGISTERING, /* called unregister_netdevice */ 1930 NETREG_UNREGISTERED, /* completed unregister todo */ 1931 NETREG_RELEASED, /* called free_netdev */ 1932 NETREG_DUMMY, /* dummy device for NAPI poll */ 1933 } reg_state:8; 1934 1935 bool dismantle; 1936 1937 enum { 1938 RTNL_LINK_INITIALIZED, 1939 RTNL_LINK_INITIALIZING, 1940 } rtnl_link_state:16; 1941 1942 bool needs_free_netdev; 1943 void (*priv_destructor)(struct net_device *dev); 1944 1945 #ifdef CONFIG_NETPOLL 1946 struct netpoll_info __rcu *npinfo; 1947 #endif 1948 1949 possible_net_t nd_net; 1950 1951 /* mid-layer private */ 1952 union { 1953 void *ml_priv; 1954 struct pcpu_lstats __percpu *lstats; 1955 struct pcpu_sw_netstats __percpu *tstats; 1956 struct pcpu_dstats __percpu *dstats; 1957 struct pcpu_vstats __percpu *vstats; 1958 }; 1959 1960 #if IS_ENABLED(CONFIG_GARP) 1961 struct garp_port __rcu *garp_port; 1962 #endif 1963 #if IS_ENABLED(CONFIG_MRP) 1964 struct mrp_port __rcu *mrp_port; 1965 #endif 1966 1967 struct device dev; 1968 const struct attribute_group *sysfs_groups[4]; 1969 const struct attribute_group *sysfs_rx_queue_group; 1970 1971 const struct rtnl_link_ops *rtnl_link_ops; 1972 1973 /* for setting kernel sock attribute on TCP connection setup */ 1974 #define GSO_MAX_SIZE 65536 1975 unsigned int gso_max_size; 1976 #define GSO_MAX_SEGS 65535 1977 u16 gso_max_segs; 1978 1979 #ifdef CONFIG_DCB 1980 const struct dcbnl_rtnl_ops *dcbnl_ops; 1981 #endif 1982 u8 num_tc; 1983 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE]; 1984 u8 prio_tc_map[TC_BITMASK + 1]; 1985 1986 #if IS_ENABLED(CONFIG_FCOE) 1987 unsigned int fcoe_ddp_xid; 1988 #endif 1989 #if IS_ENABLED(CONFIG_CGROUP_NET_PRIO) 1990 struct netprio_map __rcu *priomap; 1991 #endif 1992 struct phy_device *phydev; 1993 struct sfp_bus *sfp_bus; 1994 struct lock_class_key *qdisc_tx_busylock; 1995 struct lock_class_key *qdisc_running_key; 1996 bool proto_down; 1997 }; 1998 #define to_net_dev(d) container_of(d, struct net_device, dev) 1999 2000 static inline bool netif_elide_gro(const struct net_device *dev) 2001 { 2002 if (!(dev->features & NETIF_F_GRO) || dev->xdp_prog) 2003 return true; 2004 return false; 2005 } 2006 2007 #define NETDEV_ALIGN 32 2008 2009 static inline 2010 int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio) 2011 { 2012 return dev->prio_tc_map[prio & TC_BITMASK]; 2013 } 2014 2015 static inline 2016 int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc) 2017 { 2018 if (tc >= dev->num_tc) 2019 return -EINVAL; 2020 2021 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK; 2022 return 0; 2023 } 2024 2025 int netdev_txq_to_tc(struct net_device *dev, unsigned int txq); 2026 void netdev_reset_tc(struct net_device *dev); 2027 int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset); 2028 int netdev_set_num_tc(struct net_device *dev, u8 num_tc); 2029 2030 static inline 2031 int netdev_get_num_tc(struct net_device *dev) 2032 { 2033 return dev->num_tc; 2034 } 2035 2036 static inline 2037 struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev, 2038 unsigned int index) 2039 { 2040 return &dev->_tx[index]; 2041 } 2042 2043 static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev, 2044 const struct sk_buff *skb) 2045 { 2046 return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb)); 2047 } 2048 2049 static inline void netdev_for_each_tx_queue(struct net_device *dev, 2050 void (*f)(struct net_device *, 2051 struct netdev_queue *, 2052 void *), 2053 void *arg) 2054 { 2055 unsigned int i; 2056 2057 for (i = 0; i < dev->num_tx_queues; i++) 2058 f(dev, &dev->_tx[i], arg); 2059 } 2060 2061 #define netdev_lockdep_set_classes(dev) \ 2062 { \ 2063 static struct lock_class_key qdisc_tx_busylock_key; \ 2064 static struct lock_class_key qdisc_running_key; \ 2065 static struct lock_class_key qdisc_xmit_lock_key; \ 2066 static struct lock_class_key dev_addr_list_lock_key; \ 2067 unsigned int i; \ 2068 \ 2069 (dev)->qdisc_tx_busylock = &qdisc_tx_busylock_key; \ 2070 (dev)->qdisc_running_key = &qdisc_running_key; \ 2071 lockdep_set_class(&(dev)->addr_list_lock, \ 2072 &dev_addr_list_lock_key); \ 2073 for (i = 0; i < (dev)->num_tx_queues; i++) \ 2074 lockdep_set_class(&(dev)->_tx[i]._xmit_lock, \ 2075 &qdisc_xmit_lock_key); \ 2076 } 2077 2078 struct netdev_queue *netdev_pick_tx(struct net_device *dev, 2079 struct sk_buff *skb, 2080 void *accel_priv); 2081 2082 /* returns the headroom that the master device needs to take in account 2083 * when forwarding to this dev 2084 */ 2085 static inline unsigned netdev_get_fwd_headroom(struct net_device *dev) 2086 { 2087 return dev->priv_flags & IFF_PHONY_HEADROOM ? 0 : dev->needed_headroom; 2088 } 2089 2090 static inline void netdev_set_rx_headroom(struct net_device *dev, int new_hr) 2091 { 2092 if (dev->netdev_ops->ndo_set_rx_headroom) 2093 dev->netdev_ops->ndo_set_rx_headroom(dev, new_hr); 2094 } 2095 2096 /* set the device rx headroom to the dev's default */ 2097 static inline void netdev_reset_rx_headroom(struct net_device *dev) 2098 { 2099 netdev_set_rx_headroom(dev, -1); 2100 } 2101 2102 /* 2103 * Net namespace inlines 2104 */ 2105 static inline 2106 struct net *dev_net(const struct net_device *dev) 2107 { 2108 return read_pnet(&dev->nd_net); 2109 } 2110 2111 static inline 2112 void dev_net_set(struct net_device *dev, struct net *net) 2113 { 2114 write_pnet(&dev->nd_net, net); 2115 } 2116 2117 /** 2118 * netdev_priv - access network device private data 2119 * @dev: network device 2120 * 2121 * Get network device private data 2122 */ 2123 static inline void *netdev_priv(const struct net_device *dev) 2124 { 2125 return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN); 2126 } 2127 2128 /* Set the sysfs physical device reference for the network logical device 2129 * if set prior to registration will cause a symlink during initialization. 2130 */ 2131 #define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev)) 2132 2133 /* Set the sysfs device type for the network logical device to allow 2134 * fine-grained identification of different network device types. For 2135 * example Ethernet, Wireless LAN, Bluetooth, WiMAX etc. 2136 */ 2137 #define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype)) 2138 2139 /* Default NAPI poll() weight 2140 * Device drivers are strongly advised to not use bigger value 2141 */ 2142 #define NAPI_POLL_WEIGHT 64 2143 2144 /** 2145 * netif_napi_add - initialize a NAPI context 2146 * @dev: network device 2147 * @napi: NAPI context 2148 * @poll: polling function 2149 * @weight: default weight 2150 * 2151 * netif_napi_add() must be used to initialize a NAPI context prior to calling 2152 * *any* of the other NAPI-related functions. 2153 */ 2154 void netif_napi_add(struct net_device *dev, struct napi_struct *napi, 2155 int (*poll)(struct napi_struct *, int), int weight); 2156 2157 /** 2158 * netif_tx_napi_add - initialize a NAPI context 2159 * @dev: network device 2160 * @napi: NAPI context 2161 * @poll: polling function 2162 * @weight: default weight 2163 * 2164 * This variant of netif_napi_add() should be used from drivers using NAPI 2165 * to exclusively poll a TX queue. 2166 * This will avoid we add it into napi_hash[], thus polluting this hash table. 2167 */ 2168 static inline void netif_tx_napi_add(struct net_device *dev, 2169 struct napi_struct *napi, 2170 int (*poll)(struct napi_struct *, int), 2171 int weight) 2172 { 2173 set_bit(NAPI_STATE_NO_BUSY_POLL, &napi->state); 2174 netif_napi_add(dev, napi, poll, weight); 2175 } 2176 2177 /** 2178 * netif_napi_del - remove a NAPI context 2179 * @napi: NAPI context 2180 * 2181 * netif_napi_del() removes a NAPI context from the network device NAPI list 2182 */ 2183 void netif_napi_del(struct napi_struct *napi); 2184 2185 struct napi_gro_cb { 2186 /* Virtual address of skb_shinfo(skb)->frags[0].page + offset. */ 2187 void *frag0; 2188 2189 /* Length of frag0. */ 2190 unsigned int frag0_len; 2191 2192 /* This indicates where we are processing relative to skb->data. */ 2193 int data_offset; 2194 2195 /* This is non-zero if the packet cannot be merged with the new skb. */ 2196 u16 flush; 2197 2198 /* Save the IP ID here and check when we get to the transport layer */ 2199 u16 flush_id; 2200 2201 /* Number of segments aggregated. */ 2202 u16 count; 2203 2204 /* Start offset for remote checksum offload */ 2205 u16 gro_remcsum_start; 2206 2207 /* jiffies when first packet was created/queued */ 2208 unsigned long age; 2209 2210 /* Used in ipv6_gro_receive() and foo-over-udp */ 2211 u16 proto; 2212 2213 /* This is non-zero if the packet may be of the same flow. */ 2214 u8 same_flow:1; 2215 2216 /* Used in tunnel GRO receive */ 2217 u8 encap_mark:1; 2218 2219 /* GRO checksum is valid */ 2220 u8 csum_valid:1; 2221 2222 /* Number of checksums via CHECKSUM_UNNECESSARY */ 2223 u8 csum_cnt:3; 2224 2225 /* Free the skb? */ 2226 u8 free:2; 2227 #define NAPI_GRO_FREE 1 2228 #define NAPI_GRO_FREE_STOLEN_HEAD 2 2229 2230 /* Used in foo-over-udp, set in udp[46]_gro_receive */ 2231 u8 is_ipv6:1; 2232 2233 /* Used in GRE, set in fou/gue_gro_receive */ 2234 u8 is_fou:1; 2235 2236 /* Used to determine if flush_id can be ignored */ 2237 u8 is_atomic:1; 2238 2239 /* Number of gro_receive callbacks this packet already went through */ 2240 u8 recursion_counter:4; 2241 2242 /* 1 bit hole */ 2243 2244 /* used to support CHECKSUM_COMPLETE for tunneling protocols */ 2245 __wsum csum; 2246 2247 /* used in skb_gro_receive() slow path */ 2248 struct sk_buff *last; 2249 }; 2250 2251 #define NAPI_GRO_CB(skb) ((struct napi_gro_cb *)(skb)->cb) 2252 2253 #define GRO_RECURSION_LIMIT 15 2254 static inline int gro_recursion_inc_test(struct sk_buff *skb) 2255 { 2256 return ++NAPI_GRO_CB(skb)->recursion_counter == GRO_RECURSION_LIMIT; 2257 } 2258 2259 typedef struct sk_buff *(*gro_receive_t)(struct list_head *, struct sk_buff *); 2260 static inline struct sk_buff *call_gro_receive(gro_receive_t cb, 2261 struct list_head *head, 2262 struct sk_buff *skb) 2263 { 2264 if (unlikely(gro_recursion_inc_test(skb))) { 2265 NAPI_GRO_CB(skb)->flush |= 1; 2266 return NULL; 2267 } 2268 2269 return cb(head, skb); 2270 } 2271 2272 typedef struct sk_buff *(*gro_receive_sk_t)(struct sock *, struct list_head *, 2273 struct sk_buff *); 2274 static inline struct sk_buff *call_gro_receive_sk(gro_receive_sk_t cb, 2275 struct sock *sk, 2276 struct list_head *head, 2277 struct sk_buff *skb) 2278 { 2279 if (unlikely(gro_recursion_inc_test(skb))) { 2280 NAPI_GRO_CB(skb)->flush |= 1; 2281 return NULL; 2282 } 2283 2284 return cb(sk, head, skb); 2285 } 2286 2287 struct packet_type { 2288 __be16 type; /* This is really htons(ether_type). */ 2289 struct net_device *dev; /* NULL is wildcarded here */ 2290 int (*func) (struct sk_buff *, 2291 struct net_device *, 2292 struct packet_type *, 2293 struct net_device *); 2294 bool (*id_match)(struct packet_type *ptype, 2295 struct sock *sk); 2296 void *af_packet_priv; 2297 struct list_head list; 2298 }; 2299 2300 struct offload_callbacks { 2301 struct sk_buff *(*gso_segment)(struct sk_buff *skb, 2302 netdev_features_t features); 2303 struct sk_buff *(*gro_receive)(struct list_head *head, 2304 struct sk_buff *skb); 2305 int (*gro_complete)(struct sk_buff *skb, int nhoff); 2306 }; 2307 2308 struct packet_offload { 2309 __be16 type; /* This is really htons(ether_type). */ 2310 u16 priority; 2311 struct offload_callbacks callbacks; 2312 struct list_head list; 2313 }; 2314 2315 /* often modified stats are per-CPU, other are shared (netdev->stats) */ 2316 struct pcpu_sw_netstats { 2317 u64 rx_packets; 2318 u64 rx_bytes; 2319 u64 tx_packets; 2320 u64 tx_bytes; 2321 struct u64_stats_sync syncp; 2322 }; 2323 2324 #define __netdev_alloc_pcpu_stats(type, gfp) \ 2325 ({ \ 2326 typeof(type) __percpu *pcpu_stats = alloc_percpu_gfp(type, gfp);\ 2327 if (pcpu_stats) { \ 2328 int __cpu; \ 2329 for_each_possible_cpu(__cpu) { \ 2330 typeof(type) *stat; \ 2331 stat = per_cpu_ptr(pcpu_stats, __cpu); \ 2332 u64_stats_init(&stat->syncp); \ 2333 } \ 2334 } \ 2335 pcpu_stats; \ 2336 }) 2337 2338 #define netdev_alloc_pcpu_stats(type) \ 2339 __netdev_alloc_pcpu_stats(type, GFP_KERNEL) 2340 2341 enum netdev_lag_tx_type { 2342 NETDEV_LAG_TX_TYPE_UNKNOWN, 2343 NETDEV_LAG_TX_TYPE_RANDOM, 2344 NETDEV_LAG_TX_TYPE_BROADCAST, 2345 NETDEV_LAG_TX_TYPE_ROUNDROBIN, 2346 NETDEV_LAG_TX_TYPE_ACTIVEBACKUP, 2347 NETDEV_LAG_TX_TYPE_HASH, 2348 }; 2349 2350 enum netdev_lag_hash { 2351 NETDEV_LAG_HASH_NONE, 2352 NETDEV_LAG_HASH_L2, 2353 NETDEV_LAG_HASH_L34, 2354 NETDEV_LAG_HASH_L23, 2355 NETDEV_LAG_HASH_E23, 2356 NETDEV_LAG_HASH_E34, 2357 NETDEV_LAG_HASH_UNKNOWN, 2358 }; 2359 2360 struct netdev_lag_upper_info { 2361 enum netdev_lag_tx_type tx_type; 2362 enum netdev_lag_hash hash_type; 2363 }; 2364 2365 struct netdev_lag_lower_state_info { 2366 u8 link_up : 1, 2367 tx_enabled : 1; 2368 }; 2369 2370 #include <linux/notifier.h> 2371 2372 /* netdevice notifier chain. Please remember to update netdev_cmd_to_name() 2373 * and the rtnetlink notification exclusion list in rtnetlink_event() when 2374 * adding new types. 2375 */ 2376 enum netdev_cmd { 2377 NETDEV_UP = 1, /* For now you can't veto a device up/down */ 2378 NETDEV_DOWN, 2379 NETDEV_REBOOT, /* Tell a protocol stack a network interface 2380 detected a hardware crash and restarted 2381 - we can use this eg to kick tcp sessions 2382 once done */ 2383 NETDEV_CHANGE, /* Notify device state change */ 2384 NETDEV_REGISTER, 2385 NETDEV_UNREGISTER, 2386 NETDEV_CHANGEMTU, /* notify after mtu change happened */ 2387 NETDEV_CHANGEADDR, 2388 NETDEV_GOING_DOWN, 2389 NETDEV_CHANGENAME, 2390 NETDEV_FEAT_CHANGE, 2391 NETDEV_BONDING_FAILOVER, 2392 NETDEV_PRE_UP, 2393 NETDEV_PRE_TYPE_CHANGE, 2394 NETDEV_POST_TYPE_CHANGE, 2395 NETDEV_POST_INIT, 2396 NETDEV_RELEASE, 2397 NETDEV_NOTIFY_PEERS, 2398 NETDEV_JOIN, 2399 NETDEV_CHANGEUPPER, 2400 NETDEV_RESEND_IGMP, 2401 NETDEV_PRECHANGEMTU, /* notify before mtu change happened */ 2402 NETDEV_CHANGEINFODATA, 2403 NETDEV_BONDING_INFO, 2404 NETDEV_PRECHANGEUPPER, 2405 NETDEV_CHANGELOWERSTATE, 2406 NETDEV_UDP_TUNNEL_PUSH_INFO, 2407 NETDEV_UDP_TUNNEL_DROP_INFO, 2408 NETDEV_CHANGE_TX_QUEUE_LEN, 2409 NETDEV_CVLAN_FILTER_PUSH_INFO, 2410 NETDEV_CVLAN_FILTER_DROP_INFO, 2411 NETDEV_SVLAN_FILTER_PUSH_INFO, 2412 NETDEV_SVLAN_FILTER_DROP_INFO, 2413 }; 2414 const char *netdev_cmd_to_name(enum netdev_cmd cmd); 2415 2416 int register_netdevice_notifier(struct notifier_block *nb); 2417 int unregister_netdevice_notifier(struct notifier_block *nb); 2418 2419 struct netdev_notifier_info { 2420 struct net_device *dev; 2421 struct netlink_ext_ack *extack; 2422 }; 2423 2424 struct netdev_notifier_change_info { 2425 struct netdev_notifier_info info; /* must be first */ 2426 unsigned int flags_changed; 2427 }; 2428 2429 struct netdev_notifier_changeupper_info { 2430 struct netdev_notifier_info info; /* must be first */ 2431 struct net_device *upper_dev; /* new upper dev */ 2432 bool master; /* is upper dev master */ 2433 bool linking; /* is the notification for link or unlink */ 2434 void *upper_info; /* upper dev info */ 2435 }; 2436 2437 struct netdev_notifier_changelowerstate_info { 2438 struct netdev_notifier_info info; /* must be first */ 2439 void *lower_state_info; /* is lower dev state */ 2440 }; 2441 2442 static inline void netdev_notifier_info_init(struct netdev_notifier_info *info, 2443 struct net_device *dev) 2444 { 2445 info->dev = dev; 2446 info->extack = NULL; 2447 } 2448 2449 static inline struct net_device * 2450 netdev_notifier_info_to_dev(const struct netdev_notifier_info *info) 2451 { 2452 return info->dev; 2453 } 2454 2455 static inline struct netlink_ext_ack * 2456 netdev_notifier_info_to_extack(const struct netdev_notifier_info *info) 2457 { 2458 return info->extack; 2459 } 2460 2461 int call_netdevice_notifiers(unsigned long val, struct net_device *dev); 2462 2463 2464 extern rwlock_t dev_base_lock; /* Device list lock */ 2465 2466 #define for_each_netdev(net, d) \ 2467 list_for_each_entry(d, &(net)->dev_base_head, dev_list) 2468 #define for_each_netdev_reverse(net, d) \ 2469 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list) 2470 #define for_each_netdev_rcu(net, d) \ 2471 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list) 2472 #define for_each_netdev_safe(net, d, n) \ 2473 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list) 2474 #define for_each_netdev_continue(net, d) \ 2475 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list) 2476 #define for_each_netdev_continue_rcu(net, d) \ 2477 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list) 2478 #define for_each_netdev_in_bond_rcu(bond, slave) \ 2479 for_each_netdev_rcu(&init_net, slave) \ 2480 if (netdev_master_upper_dev_get_rcu(slave) == (bond)) 2481 #define net_device_entry(lh) list_entry(lh, struct net_device, dev_list) 2482 2483 static inline struct net_device *next_net_device(struct net_device *dev) 2484 { 2485 struct list_head *lh; 2486 struct net *net; 2487 2488 net = dev_net(dev); 2489 lh = dev->dev_list.next; 2490 return lh == &net->dev_base_head ? NULL : net_device_entry(lh); 2491 } 2492 2493 static inline struct net_device *next_net_device_rcu(struct net_device *dev) 2494 { 2495 struct list_head *lh; 2496 struct net *net; 2497 2498 net = dev_net(dev); 2499 lh = rcu_dereference(list_next_rcu(&dev->dev_list)); 2500 return lh == &net->dev_base_head ? NULL : net_device_entry(lh); 2501 } 2502 2503 static inline struct net_device *first_net_device(struct net *net) 2504 { 2505 return list_empty(&net->dev_base_head) ? NULL : 2506 net_device_entry(net->dev_base_head.next); 2507 } 2508 2509 static inline struct net_device *first_net_device_rcu(struct net *net) 2510 { 2511 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head)); 2512 2513 return lh == &net->dev_base_head ? NULL : net_device_entry(lh); 2514 } 2515 2516 int netdev_boot_setup_check(struct net_device *dev); 2517 unsigned long netdev_boot_base(const char *prefix, int unit); 2518 struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type, 2519 const char *hwaddr); 2520 struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type); 2521 struct net_device *__dev_getfirstbyhwtype(struct net *net, unsigned short type); 2522 void dev_add_pack(struct packet_type *pt); 2523 void dev_remove_pack(struct packet_type *pt); 2524 void __dev_remove_pack(struct packet_type *pt); 2525 void dev_add_offload(struct packet_offload *po); 2526 void dev_remove_offload(struct packet_offload *po); 2527 2528 int dev_get_iflink(const struct net_device *dev); 2529 int dev_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb); 2530 struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags, 2531 unsigned short mask); 2532 struct net_device *dev_get_by_name(struct net *net, const char *name); 2533 struct net_device *dev_get_by_name_rcu(struct net *net, const char *name); 2534 struct net_device *__dev_get_by_name(struct net *net, const char *name); 2535 int dev_alloc_name(struct net_device *dev, const char *name); 2536 int dev_open(struct net_device *dev); 2537 void dev_close(struct net_device *dev); 2538 void dev_close_many(struct list_head *head, bool unlink); 2539 void dev_disable_lro(struct net_device *dev); 2540 int dev_loopback_xmit(struct net *net, struct sock *sk, struct sk_buff *newskb); 2541 int dev_queue_xmit(struct sk_buff *skb); 2542 int dev_queue_xmit_accel(struct sk_buff *skb, void *accel_priv); 2543 int dev_direct_xmit(struct sk_buff *skb, u16 queue_id); 2544 int register_netdevice(struct net_device *dev); 2545 void unregister_netdevice_queue(struct net_device *dev, struct list_head *head); 2546 void unregister_netdevice_many(struct list_head *head); 2547 static inline void unregister_netdevice(struct net_device *dev) 2548 { 2549 unregister_netdevice_queue(dev, NULL); 2550 } 2551 2552 int netdev_refcnt_read(const struct net_device *dev); 2553 void free_netdev(struct net_device *dev); 2554 void netdev_freemem(struct net_device *dev); 2555 void synchronize_net(void); 2556 int init_dummy_netdev(struct net_device *dev); 2557 2558 DECLARE_PER_CPU(int, xmit_recursion); 2559 #define XMIT_RECURSION_LIMIT 10 2560 2561 static inline int dev_recursion_level(void) 2562 { 2563 return this_cpu_read(xmit_recursion); 2564 } 2565 2566 struct net_device *dev_get_by_index(struct net *net, int ifindex); 2567 struct net_device *__dev_get_by_index(struct net *net, int ifindex); 2568 struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex); 2569 struct net_device *dev_get_by_napi_id(unsigned int napi_id); 2570 int netdev_get_name(struct net *net, char *name, int ifindex); 2571 int dev_restart(struct net_device *dev); 2572 int skb_gro_receive(struct sk_buff *p, struct sk_buff *skb); 2573 2574 static inline unsigned int skb_gro_offset(const struct sk_buff *skb) 2575 { 2576 return NAPI_GRO_CB(skb)->data_offset; 2577 } 2578 2579 static inline unsigned int skb_gro_len(const struct sk_buff *skb) 2580 { 2581 return skb->len - NAPI_GRO_CB(skb)->data_offset; 2582 } 2583 2584 static inline void skb_gro_pull(struct sk_buff *skb, unsigned int len) 2585 { 2586 NAPI_GRO_CB(skb)->data_offset += len; 2587 } 2588 2589 static inline void *skb_gro_header_fast(struct sk_buff *skb, 2590 unsigned int offset) 2591 { 2592 return NAPI_GRO_CB(skb)->frag0 + offset; 2593 } 2594 2595 static inline int skb_gro_header_hard(struct sk_buff *skb, unsigned int hlen) 2596 { 2597 return NAPI_GRO_CB(skb)->frag0_len < hlen; 2598 } 2599 2600 static inline void skb_gro_frag0_invalidate(struct sk_buff *skb) 2601 { 2602 NAPI_GRO_CB(skb)->frag0 = NULL; 2603 NAPI_GRO_CB(skb)->frag0_len = 0; 2604 } 2605 2606 static inline void *skb_gro_header_slow(struct sk_buff *skb, unsigned int hlen, 2607 unsigned int offset) 2608 { 2609 if (!pskb_may_pull(skb, hlen)) 2610 return NULL; 2611 2612 skb_gro_frag0_invalidate(skb); 2613 return skb->data + offset; 2614 } 2615 2616 static inline void *skb_gro_network_header(struct sk_buff *skb) 2617 { 2618 return (NAPI_GRO_CB(skb)->frag0 ?: skb->data) + 2619 skb_network_offset(skb); 2620 } 2621 2622 static inline void skb_gro_postpull_rcsum(struct sk_buff *skb, 2623 const void *start, unsigned int len) 2624 { 2625 if (NAPI_GRO_CB(skb)->csum_valid) 2626 NAPI_GRO_CB(skb)->csum = csum_sub(NAPI_GRO_CB(skb)->csum, 2627 csum_partial(start, len, 0)); 2628 } 2629 2630 /* GRO checksum functions. These are logical equivalents of the normal 2631 * checksum functions (in skbuff.h) except that they operate on the GRO 2632 * offsets and fields in sk_buff. 2633 */ 2634 2635 __sum16 __skb_gro_checksum_complete(struct sk_buff *skb); 2636 2637 static inline bool skb_at_gro_remcsum_start(struct sk_buff *skb) 2638 { 2639 return (NAPI_GRO_CB(skb)->gro_remcsum_start == skb_gro_offset(skb)); 2640 } 2641 2642 static inline bool __skb_gro_checksum_validate_needed(struct sk_buff *skb, 2643 bool zero_okay, 2644 __sum16 check) 2645 { 2646 return ((skb->ip_summed != CHECKSUM_PARTIAL || 2647 skb_checksum_start_offset(skb) < 2648 skb_gro_offset(skb)) && 2649 !skb_at_gro_remcsum_start(skb) && 2650 NAPI_GRO_CB(skb)->csum_cnt == 0 && 2651 (!zero_okay || check)); 2652 } 2653 2654 static inline __sum16 __skb_gro_checksum_validate_complete(struct sk_buff *skb, 2655 __wsum psum) 2656 { 2657 if (NAPI_GRO_CB(skb)->csum_valid && 2658 !csum_fold(csum_add(psum, NAPI_GRO_CB(skb)->csum))) 2659 return 0; 2660 2661 NAPI_GRO_CB(skb)->csum = psum; 2662 2663 return __skb_gro_checksum_complete(skb); 2664 } 2665 2666 static inline void skb_gro_incr_csum_unnecessary(struct sk_buff *skb) 2667 { 2668 if (NAPI_GRO_CB(skb)->csum_cnt > 0) { 2669 /* Consume a checksum from CHECKSUM_UNNECESSARY */ 2670 NAPI_GRO_CB(skb)->csum_cnt--; 2671 } else { 2672 /* Update skb for CHECKSUM_UNNECESSARY and csum_level when we 2673 * verified a new top level checksum or an encapsulated one 2674 * during GRO. This saves work if we fallback to normal path. 2675 */ 2676 __skb_incr_checksum_unnecessary(skb); 2677 } 2678 } 2679 2680 #define __skb_gro_checksum_validate(skb, proto, zero_okay, check, \ 2681 compute_pseudo) \ 2682 ({ \ 2683 __sum16 __ret = 0; \ 2684 if (__skb_gro_checksum_validate_needed(skb, zero_okay, check)) \ 2685 __ret = __skb_gro_checksum_validate_complete(skb, \ 2686 compute_pseudo(skb, proto)); \ 2687 if (!__ret) \ 2688 skb_gro_incr_csum_unnecessary(skb); \ 2689 __ret; \ 2690 }) 2691 2692 #define skb_gro_checksum_validate(skb, proto, compute_pseudo) \ 2693 __skb_gro_checksum_validate(skb, proto, false, 0, compute_pseudo) 2694 2695 #define skb_gro_checksum_validate_zero_check(skb, proto, check, \ 2696 compute_pseudo) \ 2697 __skb_gro_checksum_validate(skb, proto, true, check, compute_pseudo) 2698 2699 #define skb_gro_checksum_simple_validate(skb) \ 2700 __skb_gro_checksum_validate(skb, 0, false, 0, null_compute_pseudo) 2701 2702 static inline bool __skb_gro_checksum_convert_check(struct sk_buff *skb) 2703 { 2704 return (NAPI_GRO_CB(skb)->csum_cnt == 0 && 2705 !NAPI_GRO_CB(skb)->csum_valid); 2706 } 2707 2708 static inline void __skb_gro_checksum_convert(struct sk_buff *skb, 2709 __sum16 check, __wsum pseudo) 2710 { 2711 NAPI_GRO_CB(skb)->csum = ~pseudo; 2712 NAPI_GRO_CB(skb)->csum_valid = 1; 2713 } 2714 2715 #define skb_gro_checksum_try_convert(skb, proto, check, compute_pseudo) \ 2716 do { \ 2717 if (__skb_gro_checksum_convert_check(skb)) \ 2718 __skb_gro_checksum_convert(skb, check, \ 2719 compute_pseudo(skb, proto)); \ 2720 } while (0) 2721 2722 struct gro_remcsum { 2723 int offset; 2724 __wsum delta; 2725 }; 2726 2727 static inline void skb_gro_remcsum_init(struct gro_remcsum *grc) 2728 { 2729 grc->offset = 0; 2730 grc->delta = 0; 2731 } 2732 2733 static inline void *skb_gro_remcsum_process(struct sk_buff *skb, void *ptr, 2734 unsigned int off, size_t hdrlen, 2735 int start, int offset, 2736 struct gro_remcsum *grc, 2737 bool nopartial) 2738 { 2739 __wsum delta; 2740 size_t plen = hdrlen + max_t(size_t, offset + sizeof(u16), start); 2741 2742 BUG_ON(!NAPI_GRO_CB(skb)->csum_valid); 2743 2744 if (!nopartial) { 2745 NAPI_GRO_CB(skb)->gro_remcsum_start = off + hdrlen + start; 2746 return ptr; 2747 } 2748 2749 ptr = skb_gro_header_fast(skb, off); 2750 if (skb_gro_header_hard(skb, off + plen)) { 2751 ptr = skb_gro_header_slow(skb, off + plen, off); 2752 if (!ptr) 2753 return NULL; 2754 } 2755 2756 delta = remcsum_adjust(ptr + hdrlen, NAPI_GRO_CB(skb)->csum, 2757 start, offset); 2758 2759 /* Adjust skb->csum since we changed the packet */ 2760 NAPI_GRO_CB(skb)->csum = csum_add(NAPI_GRO_CB(skb)->csum, delta); 2761 2762 grc->offset = off + hdrlen + offset; 2763 grc->delta = delta; 2764 2765 return ptr; 2766 } 2767 2768 static inline void skb_gro_remcsum_cleanup(struct sk_buff *skb, 2769 struct gro_remcsum *grc) 2770 { 2771 void *ptr; 2772 size_t plen = grc->offset + sizeof(u16); 2773 2774 if (!grc->delta) 2775 return; 2776 2777 ptr = skb_gro_header_fast(skb, grc->offset); 2778 if (skb_gro_header_hard(skb, grc->offset + sizeof(u16))) { 2779 ptr = skb_gro_header_slow(skb, plen, grc->offset); 2780 if (!ptr) 2781 return; 2782 } 2783 2784 remcsum_unadjust((__sum16 *)ptr, grc->delta); 2785 } 2786 2787 #ifdef CONFIG_XFRM_OFFLOAD 2788 static inline void skb_gro_flush_final(struct sk_buff *skb, struct sk_buff *pp, int flush) 2789 { 2790 if (PTR_ERR(pp) != -EINPROGRESS) 2791 NAPI_GRO_CB(skb)->flush |= flush; 2792 } 2793 #else 2794 static inline void skb_gro_flush_final(struct sk_buff *skb, struct sk_buff *pp, int flush) 2795 { 2796 NAPI_GRO_CB(skb)->flush |= flush; 2797 } 2798 #endif 2799 2800 static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev, 2801 unsigned short type, 2802 const void *daddr, const void *saddr, 2803 unsigned int len) 2804 { 2805 if (!dev->header_ops || !dev->header_ops->create) 2806 return 0; 2807 2808 return dev->header_ops->create(skb, dev, type, daddr, saddr, len); 2809 } 2810 2811 static inline int dev_parse_header(const struct sk_buff *skb, 2812 unsigned char *haddr) 2813 { 2814 const struct net_device *dev = skb->dev; 2815 2816 if (!dev->header_ops || !dev->header_ops->parse) 2817 return 0; 2818 return dev->header_ops->parse(skb, haddr); 2819 } 2820 2821 /* ll_header must have at least hard_header_len allocated */ 2822 static inline bool dev_validate_header(const struct net_device *dev, 2823 char *ll_header, int len) 2824 { 2825 if (likely(len >= dev->hard_header_len)) 2826 return true; 2827 if (len < dev->min_header_len) 2828 return false; 2829 2830 if (capable(CAP_SYS_RAWIO)) { 2831 memset(ll_header + len, 0, dev->hard_header_len - len); 2832 return true; 2833 } 2834 2835 if (dev->header_ops && dev->header_ops->validate) 2836 return dev->header_ops->validate(ll_header, len); 2837 2838 return false; 2839 } 2840 2841 typedef int gifconf_func_t(struct net_device * dev, char __user * bufptr, 2842 int len, int size); 2843 int register_gifconf(unsigned int family, gifconf_func_t *gifconf); 2844 static inline int unregister_gifconf(unsigned int family) 2845 { 2846 return register_gifconf(family, NULL); 2847 } 2848 2849 #ifdef CONFIG_NET_FLOW_LIMIT 2850 #define FLOW_LIMIT_HISTORY (1 << 7) /* must be ^2 and !overflow buckets */ 2851 struct sd_flow_limit { 2852 u64 count; 2853 unsigned int num_buckets; 2854 unsigned int history_head; 2855 u16 history[FLOW_LIMIT_HISTORY]; 2856 u8 buckets[]; 2857 }; 2858 2859 extern int netdev_flow_limit_table_len; 2860 #endif /* CONFIG_NET_FLOW_LIMIT */ 2861 2862 /* 2863 * Incoming packets are placed on per-CPU queues 2864 */ 2865 struct softnet_data { 2866 struct list_head poll_list; 2867 struct sk_buff_head process_queue; 2868 2869 /* stats */ 2870 unsigned int processed; 2871 unsigned int time_squeeze; 2872 unsigned int received_rps; 2873 #ifdef CONFIG_RPS 2874 struct softnet_data *rps_ipi_list; 2875 #endif 2876 #ifdef CONFIG_NET_FLOW_LIMIT 2877 struct sd_flow_limit __rcu *flow_limit; 2878 #endif 2879 struct Qdisc *output_queue; 2880 struct Qdisc **output_queue_tailp; 2881 struct sk_buff *completion_queue; 2882 #ifdef CONFIG_XFRM_OFFLOAD 2883 struct sk_buff_head xfrm_backlog; 2884 #endif 2885 #ifdef CONFIG_RPS 2886 /* input_queue_head should be written by cpu owning this struct, 2887 * and only read by other cpus. Worth using a cache line. 2888 */ 2889 unsigned int input_queue_head ____cacheline_aligned_in_smp; 2890 2891 /* Elements below can be accessed between CPUs for RPS/RFS */ 2892 call_single_data_t csd ____cacheline_aligned_in_smp; 2893 struct softnet_data *rps_ipi_next; 2894 unsigned int cpu; 2895 unsigned int input_queue_tail; 2896 #endif 2897 unsigned int dropped; 2898 struct sk_buff_head input_pkt_queue; 2899 struct napi_struct backlog; 2900 2901 }; 2902 2903 static inline void input_queue_head_incr(struct softnet_data *sd) 2904 { 2905 #ifdef CONFIG_RPS 2906 sd->input_queue_head++; 2907 #endif 2908 } 2909 2910 static inline void input_queue_tail_incr_save(struct softnet_data *sd, 2911 unsigned int *qtail) 2912 { 2913 #ifdef CONFIG_RPS 2914 *qtail = ++sd->input_queue_tail; 2915 #endif 2916 } 2917 2918 DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data); 2919 2920 void __netif_schedule(struct Qdisc *q); 2921 void netif_schedule_queue(struct netdev_queue *txq); 2922 2923 static inline void netif_tx_schedule_all(struct net_device *dev) 2924 { 2925 unsigned int i; 2926 2927 for (i = 0; i < dev->num_tx_queues; i++) 2928 netif_schedule_queue(netdev_get_tx_queue(dev, i)); 2929 } 2930 2931 static __always_inline void netif_tx_start_queue(struct netdev_queue *dev_queue) 2932 { 2933 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state); 2934 } 2935 2936 /** 2937 * netif_start_queue - allow transmit 2938 * @dev: network device 2939 * 2940 * Allow upper layers to call the device hard_start_xmit routine. 2941 */ 2942 static inline void netif_start_queue(struct net_device *dev) 2943 { 2944 netif_tx_start_queue(netdev_get_tx_queue(dev, 0)); 2945 } 2946 2947 static inline void netif_tx_start_all_queues(struct net_device *dev) 2948 { 2949 unsigned int i; 2950 2951 for (i = 0; i < dev->num_tx_queues; i++) { 2952 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 2953 netif_tx_start_queue(txq); 2954 } 2955 } 2956 2957 void netif_tx_wake_queue(struct netdev_queue *dev_queue); 2958 2959 /** 2960 * netif_wake_queue - restart transmit 2961 * @dev: network device 2962 * 2963 * Allow upper layers to call the device hard_start_xmit routine. 2964 * Used for flow control when transmit resources are available. 2965 */ 2966 static inline void netif_wake_queue(struct net_device *dev) 2967 { 2968 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0)); 2969 } 2970 2971 static inline void netif_tx_wake_all_queues(struct net_device *dev) 2972 { 2973 unsigned int i; 2974 2975 for (i = 0; i < dev->num_tx_queues; i++) { 2976 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 2977 netif_tx_wake_queue(txq); 2978 } 2979 } 2980 2981 static __always_inline void netif_tx_stop_queue(struct netdev_queue *dev_queue) 2982 { 2983 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state); 2984 } 2985 2986 /** 2987 * netif_stop_queue - stop transmitted packets 2988 * @dev: network device 2989 * 2990 * Stop upper layers calling the device hard_start_xmit routine. 2991 * Used for flow control when transmit resources are unavailable. 2992 */ 2993 static inline void netif_stop_queue(struct net_device *dev) 2994 { 2995 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0)); 2996 } 2997 2998 void netif_tx_stop_all_queues(struct net_device *dev); 2999 3000 static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue) 3001 { 3002 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state); 3003 } 3004 3005 /** 3006 * netif_queue_stopped - test if transmit queue is flowblocked 3007 * @dev: network device 3008 * 3009 * Test if transmit queue on device is currently unable to send. 3010 */ 3011 static inline bool netif_queue_stopped(const struct net_device *dev) 3012 { 3013 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0)); 3014 } 3015 3016 static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue) 3017 { 3018 return dev_queue->state & QUEUE_STATE_ANY_XOFF; 3019 } 3020 3021 static inline bool 3022 netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue) 3023 { 3024 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN; 3025 } 3026 3027 static inline bool 3028 netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue) 3029 { 3030 return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN; 3031 } 3032 3033 /** 3034 * netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write 3035 * @dev_queue: pointer to transmit queue 3036 * 3037 * BQL enabled drivers might use this helper in their ndo_start_xmit(), 3038 * to give appropriate hint to the CPU. 3039 */ 3040 static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue) 3041 { 3042 #ifdef CONFIG_BQL 3043 prefetchw(&dev_queue->dql.num_queued); 3044 #endif 3045 } 3046 3047 /** 3048 * netdev_txq_bql_complete_prefetchw - prefetch bql data for write 3049 * @dev_queue: pointer to transmit queue 3050 * 3051 * BQL enabled drivers might use this helper in their TX completion path, 3052 * to give appropriate hint to the CPU. 3053 */ 3054 static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue) 3055 { 3056 #ifdef CONFIG_BQL 3057 prefetchw(&dev_queue->dql.limit); 3058 #endif 3059 } 3060 3061 static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue, 3062 unsigned int bytes) 3063 { 3064 #ifdef CONFIG_BQL 3065 dql_queued(&dev_queue->dql, bytes); 3066 3067 if (likely(dql_avail(&dev_queue->dql) >= 0)) 3068 return; 3069 3070 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state); 3071 3072 /* 3073 * The XOFF flag must be set before checking the dql_avail below, 3074 * because in netdev_tx_completed_queue we update the dql_completed 3075 * before checking the XOFF flag. 3076 */ 3077 smp_mb(); 3078 3079 /* check again in case another CPU has just made room avail */ 3080 if (unlikely(dql_avail(&dev_queue->dql) >= 0)) 3081 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state); 3082 #endif 3083 } 3084 3085 /** 3086 * netdev_sent_queue - report the number of bytes queued to hardware 3087 * @dev: network device 3088 * @bytes: number of bytes queued to the hardware device queue 3089 * 3090 * Report the number of bytes queued for sending/completion to the network 3091 * device hardware queue. @bytes should be a good approximation and should 3092 * exactly match netdev_completed_queue() @bytes 3093 */ 3094 static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes) 3095 { 3096 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes); 3097 } 3098 3099 static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue, 3100 unsigned int pkts, unsigned int bytes) 3101 { 3102 #ifdef CONFIG_BQL 3103 if (unlikely(!bytes)) 3104 return; 3105 3106 dql_completed(&dev_queue->dql, bytes); 3107 3108 /* 3109 * Without the memory barrier there is a small possiblity that 3110 * netdev_tx_sent_queue will miss the update and cause the queue to 3111 * be stopped forever 3112 */ 3113 smp_mb(); 3114 3115 if (dql_avail(&dev_queue->dql) < 0) 3116 return; 3117 3118 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state)) 3119 netif_schedule_queue(dev_queue); 3120 #endif 3121 } 3122 3123 /** 3124 * netdev_completed_queue - report bytes and packets completed by device 3125 * @dev: network device 3126 * @pkts: actual number of packets sent over the medium 3127 * @bytes: actual number of bytes sent over the medium 3128 * 3129 * Report the number of bytes and packets transmitted by the network device 3130 * hardware queue over the physical medium, @bytes must exactly match the 3131 * @bytes amount passed to netdev_sent_queue() 3132 */ 3133 static inline void netdev_completed_queue(struct net_device *dev, 3134 unsigned int pkts, unsigned int bytes) 3135 { 3136 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes); 3137 } 3138 3139 static inline void netdev_tx_reset_queue(struct netdev_queue *q) 3140 { 3141 #ifdef CONFIG_BQL 3142 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state); 3143 dql_reset(&q->dql); 3144 #endif 3145 } 3146 3147 /** 3148 * netdev_reset_queue - reset the packets and bytes count of a network device 3149 * @dev_queue: network device 3150 * 3151 * Reset the bytes and packet count of a network device and clear the 3152 * software flow control OFF bit for this network device 3153 */ 3154 static inline void netdev_reset_queue(struct net_device *dev_queue) 3155 { 3156 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0)); 3157 } 3158 3159 /** 3160 * netdev_cap_txqueue - check if selected tx queue exceeds device queues 3161 * @dev: network device 3162 * @queue_index: given tx queue index 3163 * 3164 * Returns 0 if given tx queue index >= number of device tx queues, 3165 * otherwise returns the originally passed tx queue index. 3166 */ 3167 static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index) 3168 { 3169 if (unlikely(queue_index >= dev->real_num_tx_queues)) { 3170 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n", 3171 dev->name, queue_index, 3172 dev->real_num_tx_queues); 3173 return 0; 3174 } 3175 3176 return queue_index; 3177 } 3178 3179 /** 3180 * netif_running - test if up 3181 * @dev: network device 3182 * 3183 * Test if the device has been brought up. 3184 */ 3185 static inline bool netif_running(const struct net_device *dev) 3186 { 3187 return test_bit(__LINK_STATE_START, &dev->state); 3188 } 3189 3190 /* 3191 * Routines to manage the subqueues on a device. We only need start, 3192 * stop, and a check if it's stopped. All other device management is 3193 * done at the overall netdevice level. 3194 * Also test the device if we're multiqueue. 3195 */ 3196 3197 /** 3198 * netif_start_subqueue - allow sending packets on subqueue 3199 * @dev: network device 3200 * @queue_index: sub queue index 3201 * 3202 * Start individual transmit queue of a device with multiple transmit queues. 3203 */ 3204 static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index) 3205 { 3206 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 3207 3208 netif_tx_start_queue(txq); 3209 } 3210 3211 /** 3212 * netif_stop_subqueue - stop sending packets on subqueue 3213 * @dev: network device 3214 * @queue_index: sub queue index 3215 * 3216 * Stop individual transmit queue of a device with multiple transmit queues. 3217 */ 3218 static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index) 3219 { 3220 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 3221 netif_tx_stop_queue(txq); 3222 } 3223 3224 /** 3225 * netif_subqueue_stopped - test status of subqueue 3226 * @dev: network device 3227 * @queue_index: sub queue index 3228 * 3229 * Check individual transmit queue of a device with multiple transmit queues. 3230 */ 3231 static inline bool __netif_subqueue_stopped(const struct net_device *dev, 3232 u16 queue_index) 3233 { 3234 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 3235 3236 return netif_tx_queue_stopped(txq); 3237 } 3238 3239 static inline bool netif_subqueue_stopped(const struct net_device *dev, 3240 struct sk_buff *skb) 3241 { 3242 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb)); 3243 } 3244 3245 /** 3246 * netif_wake_subqueue - allow sending packets on subqueue 3247 * @dev: network device 3248 * @queue_index: sub queue index 3249 * 3250 * Resume individual transmit queue of a device with multiple transmit queues. 3251 */ 3252 static inline void netif_wake_subqueue(struct net_device *dev, u16 queue_index) 3253 { 3254 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 3255 3256 netif_tx_wake_queue(txq); 3257 } 3258 3259 #ifdef CONFIG_XPS 3260 int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask, 3261 u16 index); 3262 #else 3263 static inline int netif_set_xps_queue(struct net_device *dev, 3264 const struct cpumask *mask, 3265 u16 index) 3266 { 3267 return 0; 3268 } 3269 #endif 3270 3271 /** 3272 * netif_is_multiqueue - test if device has multiple transmit queues 3273 * @dev: network device 3274 * 3275 * Check if device has multiple transmit queues 3276 */ 3277 static inline bool netif_is_multiqueue(const struct net_device *dev) 3278 { 3279 return dev->num_tx_queues > 1; 3280 } 3281 3282 int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq); 3283 3284 #ifdef CONFIG_SYSFS 3285 int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq); 3286 #else 3287 static inline int netif_set_real_num_rx_queues(struct net_device *dev, 3288 unsigned int rxq) 3289 { 3290 return 0; 3291 } 3292 #endif 3293 3294 static inline struct netdev_rx_queue * 3295 __netif_get_rx_queue(struct net_device *dev, unsigned int rxq) 3296 { 3297 return dev->_rx + rxq; 3298 } 3299 3300 #ifdef CONFIG_SYSFS 3301 static inline unsigned int get_netdev_rx_queue_index( 3302 struct netdev_rx_queue *queue) 3303 { 3304 struct net_device *dev = queue->dev; 3305 int index = queue - dev->_rx; 3306 3307 BUG_ON(index >= dev->num_rx_queues); 3308 return index; 3309 } 3310 #endif 3311 3312 #define DEFAULT_MAX_NUM_RSS_QUEUES (8) 3313 int netif_get_num_default_rss_queues(void); 3314 3315 enum skb_free_reason { 3316 SKB_REASON_CONSUMED, 3317 SKB_REASON_DROPPED, 3318 }; 3319 3320 void __dev_kfree_skb_irq(struct sk_buff *skb, enum skb_free_reason reason); 3321 void __dev_kfree_skb_any(struct sk_buff *skb, enum skb_free_reason reason); 3322 3323 /* 3324 * It is not allowed to call kfree_skb() or consume_skb() from hardware 3325 * interrupt context or with hardware interrupts being disabled. 3326 * (in_irq() || irqs_disabled()) 3327 * 3328 * We provide four helpers that can be used in following contexts : 3329 * 3330 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context, 3331 * replacing kfree_skb(skb) 3332 * 3333 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context. 3334 * Typically used in place of consume_skb(skb) in TX completion path 3335 * 3336 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context, 3337 * replacing kfree_skb(skb) 3338 * 3339 * dev_consume_skb_any(skb) when caller doesn't know its current irq context, 3340 * and consumed a packet. Used in place of consume_skb(skb) 3341 */ 3342 static inline void dev_kfree_skb_irq(struct sk_buff *skb) 3343 { 3344 __dev_kfree_skb_irq(skb, SKB_REASON_DROPPED); 3345 } 3346 3347 static inline void dev_consume_skb_irq(struct sk_buff *skb) 3348 { 3349 __dev_kfree_skb_irq(skb, SKB_REASON_CONSUMED); 3350 } 3351 3352 static inline void dev_kfree_skb_any(struct sk_buff *skb) 3353 { 3354 __dev_kfree_skb_any(skb, SKB_REASON_DROPPED); 3355 } 3356 3357 static inline void dev_consume_skb_any(struct sk_buff *skb) 3358 { 3359 __dev_kfree_skb_any(skb, SKB_REASON_CONSUMED); 3360 } 3361 3362 void generic_xdp_tx(struct sk_buff *skb, struct bpf_prog *xdp_prog); 3363 int do_xdp_generic(struct bpf_prog *xdp_prog, struct sk_buff *skb); 3364 int netif_rx(struct sk_buff *skb); 3365 int netif_rx_ni(struct sk_buff *skb); 3366 int netif_receive_skb(struct sk_buff *skb); 3367 int netif_receive_skb_core(struct sk_buff *skb); 3368 gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb); 3369 void napi_gro_flush(struct napi_struct *napi, bool flush_old); 3370 struct sk_buff *napi_get_frags(struct napi_struct *napi); 3371 gro_result_t napi_gro_frags(struct napi_struct *napi); 3372 struct packet_offload *gro_find_receive_by_type(__be16 type); 3373 struct packet_offload *gro_find_complete_by_type(__be16 type); 3374 3375 static inline void napi_free_frags(struct napi_struct *napi) 3376 { 3377 kfree_skb(napi->skb); 3378 napi->skb = NULL; 3379 } 3380 3381 bool netdev_is_rx_handler_busy(struct net_device *dev); 3382 int netdev_rx_handler_register(struct net_device *dev, 3383 rx_handler_func_t *rx_handler, 3384 void *rx_handler_data); 3385 void netdev_rx_handler_unregister(struct net_device *dev); 3386 3387 bool dev_valid_name(const char *name); 3388 int dev_ioctl(struct net *net, unsigned int cmd, struct ifreq *ifr, 3389 bool *need_copyout); 3390 int dev_ifconf(struct net *net, struct ifconf *, int); 3391 int dev_ethtool(struct net *net, struct ifreq *); 3392 unsigned int dev_get_flags(const struct net_device *); 3393 int __dev_change_flags(struct net_device *, unsigned int flags); 3394 int dev_change_flags(struct net_device *, unsigned int); 3395 void __dev_notify_flags(struct net_device *, unsigned int old_flags, 3396 unsigned int gchanges); 3397 int dev_change_name(struct net_device *, const char *); 3398 int dev_set_alias(struct net_device *, const char *, size_t); 3399 int dev_get_alias(const struct net_device *, char *, size_t); 3400 int dev_change_net_namespace(struct net_device *, struct net *, const char *); 3401 int __dev_set_mtu(struct net_device *, int); 3402 int dev_set_mtu(struct net_device *, int); 3403 int dev_change_tx_queue_len(struct net_device *, unsigned long); 3404 void dev_set_group(struct net_device *, int); 3405 int dev_set_mac_address(struct net_device *, struct sockaddr *); 3406 int dev_change_carrier(struct net_device *, bool new_carrier); 3407 int dev_get_phys_port_id(struct net_device *dev, 3408 struct netdev_phys_item_id *ppid); 3409 int dev_get_phys_port_name(struct net_device *dev, 3410 char *name, size_t len); 3411 int dev_change_proto_down(struct net_device *dev, bool proto_down); 3412 struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev, bool *again); 3413 struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev, 3414 struct netdev_queue *txq, int *ret); 3415 3416 typedef int (*bpf_op_t)(struct net_device *dev, struct netdev_bpf *bpf); 3417 int dev_change_xdp_fd(struct net_device *dev, struct netlink_ext_ack *extack, 3418 int fd, u32 flags); 3419 void __dev_xdp_query(struct net_device *dev, bpf_op_t xdp_op, 3420 struct netdev_bpf *xdp); 3421 3422 int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb); 3423 int dev_forward_skb(struct net_device *dev, struct sk_buff *skb); 3424 bool is_skb_forwardable(const struct net_device *dev, 3425 const struct sk_buff *skb); 3426 3427 static __always_inline int ____dev_forward_skb(struct net_device *dev, 3428 struct sk_buff *skb) 3429 { 3430 if (skb_orphan_frags(skb, GFP_ATOMIC) || 3431 unlikely(!is_skb_forwardable(dev, skb))) { 3432 atomic_long_inc(&dev->rx_dropped); 3433 kfree_skb(skb); 3434 return NET_RX_DROP; 3435 } 3436 3437 skb_scrub_packet(skb, true); 3438 skb->priority = 0; 3439 return 0; 3440 } 3441 3442 void dev_queue_xmit_nit(struct sk_buff *skb, struct net_device *dev); 3443 3444 extern int netdev_budget; 3445 extern unsigned int netdev_budget_usecs; 3446 3447 /* Called by rtnetlink.c:rtnl_unlock() */ 3448 void netdev_run_todo(void); 3449 3450 /** 3451 * dev_put - release reference to device 3452 * @dev: network device 3453 * 3454 * Release reference to device to allow it to be freed. 3455 */ 3456 static inline void dev_put(struct net_device *dev) 3457 { 3458 this_cpu_dec(*dev->pcpu_refcnt); 3459 } 3460 3461 /** 3462 * dev_hold - get reference to device 3463 * @dev: network device 3464 * 3465 * Hold reference to device to keep it from being freed. 3466 */ 3467 static inline void dev_hold(struct net_device *dev) 3468 { 3469 this_cpu_inc(*dev->pcpu_refcnt); 3470 } 3471 3472 /* Carrier loss detection, dial on demand. The functions netif_carrier_on 3473 * and _off may be called from IRQ context, but it is caller 3474 * who is responsible for serialization of these calls. 3475 * 3476 * The name carrier is inappropriate, these functions should really be 3477 * called netif_lowerlayer_*() because they represent the state of any 3478 * kind of lower layer not just hardware media. 3479 */ 3480 3481 void linkwatch_init_dev(struct net_device *dev); 3482 void linkwatch_fire_event(struct net_device *dev); 3483 void linkwatch_forget_dev(struct net_device *dev); 3484 3485 /** 3486 * netif_carrier_ok - test if carrier present 3487 * @dev: network device 3488 * 3489 * Check if carrier is present on device 3490 */ 3491 static inline bool netif_carrier_ok(const struct net_device *dev) 3492 { 3493 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state); 3494 } 3495 3496 unsigned long dev_trans_start(struct net_device *dev); 3497 3498 void __netdev_watchdog_up(struct net_device *dev); 3499 3500 void netif_carrier_on(struct net_device *dev); 3501 3502 void netif_carrier_off(struct net_device *dev); 3503 3504 /** 3505 * netif_dormant_on - mark device as dormant. 3506 * @dev: network device 3507 * 3508 * Mark device as dormant (as per RFC2863). 3509 * 3510 * The dormant state indicates that the relevant interface is not 3511 * actually in a condition to pass packets (i.e., it is not 'up') but is 3512 * in a "pending" state, waiting for some external event. For "on- 3513 * demand" interfaces, this new state identifies the situation where the 3514 * interface is waiting for events to place it in the up state. 3515 */ 3516 static inline void netif_dormant_on(struct net_device *dev) 3517 { 3518 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state)) 3519 linkwatch_fire_event(dev); 3520 } 3521 3522 /** 3523 * netif_dormant_off - set device as not dormant. 3524 * @dev: network device 3525 * 3526 * Device is not in dormant state. 3527 */ 3528 static inline void netif_dormant_off(struct net_device *dev) 3529 { 3530 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state)) 3531 linkwatch_fire_event(dev); 3532 } 3533 3534 /** 3535 * netif_dormant - test if device is dormant 3536 * @dev: network device 3537 * 3538 * Check if device is dormant. 3539 */ 3540 static inline bool netif_dormant(const struct net_device *dev) 3541 { 3542 return test_bit(__LINK_STATE_DORMANT, &dev->state); 3543 } 3544 3545 3546 /** 3547 * netif_oper_up - test if device is operational 3548 * @dev: network device 3549 * 3550 * Check if carrier is operational 3551 */ 3552 static inline bool netif_oper_up(const struct net_device *dev) 3553 { 3554 return (dev->operstate == IF_OPER_UP || 3555 dev->operstate == IF_OPER_UNKNOWN /* backward compat */); 3556 } 3557 3558 /** 3559 * netif_device_present - is device available or removed 3560 * @dev: network device 3561 * 3562 * Check if device has not been removed from system. 3563 */ 3564 static inline bool netif_device_present(struct net_device *dev) 3565 { 3566 return test_bit(__LINK_STATE_PRESENT, &dev->state); 3567 } 3568 3569 void netif_device_detach(struct net_device *dev); 3570 3571 void netif_device_attach(struct net_device *dev); 3572 3573 /* 3574 * Network interface message level settings 3575 */ 3576 3577 enum { 3578 NETIF_MSG_DRV = 0x0001, 3579 NETIF_MSG_PROBE = 0x0002, 3580 NETIF_MSG_LINK = 0x0004, 3581 NETIF_MSG_TIMER = 0x0008, 3582 NETIF_MSG_IFDOWN = 0x0010, 3583 NETIF_MSG_IFUP = 0x0020, 3584 NETIF_MSG_RX_ERR = 0x0040, 3585 NETIF_MSG_TX_ERR = 0x0080, 3586 NETIF_MSG_TX_QUEUED = 0x0100, 3587 NETIF_MSG_INTR = 0x0200, 3588 NETIF_MSG_TX_DONE = 0x0400, 3589 NETIF_MSG_RX_STATUS = 0x0800, 3590 NETIF_MSG_PKTDATA = 0x1000, 3591 NETIF_MSG_HW = 0x2000, 3592 NETIF_MSG_WOL = 0x4000, 3593 }; 3594 3595 #define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV) 3596 #define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE) 3597 #define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK) 3598 #define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER) 3599 #define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN) 3600 #define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP) 3601 #define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR) 3602 #define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR) 3603 #define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED) 3604 #define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR) 3605 #define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE) 3606 #define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS) 3607 #define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA) 3608 #define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW) 3609 #define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL) 3610 3611 static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits) 3612 { 3613 /* use default */ 3614 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8)) 3615 return default_msg_enable_bits; 3616 if (debug_value == 0) /* no output */ 3617 return 0; 3618 /* set low N bits */ 3619 return (1 << debug_value) - 1; 3620 } 3621 3622 static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu) 3623 { 3624 spin_lock(&txq->_xmit_lock); 3625 txq->xmit_lock_owner = cpu; 3626 } 3627 3628 static inline bool __netif_tx_acquire(struct netdev_queue *txq) 3629 { 3630 __acquire(&txq->_xmit_lock); 3631 return true; 3632 } 3633 3634 static inline void __netif_tx_release(struct netdev_queue *txq) 3635 { 3636 __release(&txq->_xmit_lock); 3637 } 3638 3639 static inline void __netif_tx_lock_bh(struct netdev_queue *txq) 3640 { 3641 spin_lock_bh(&txq->_xmit_lock); 3642 txq->xmit_lock_owner = smp_processor_id(); 3643 } 3644 3645 static inline bool __netif_tx_trylock(struct netdev_queue *txq) 3646 { 3647 bool ok = spin_trylock(&txq->_xmit_lock); 3648 if (likely(ok)) 3649 txq->xmit_lock_owner = smp_processor_id(); 3650 return ok; 3651 } 3652 3653 static inline void __netif_tx_unlock(struct netdev_queue *txq) 3654 { 3655 txq->xmit_lock_owner = -1; 3656 spin_unlock(&txq->_xmit_lock); 3657 } 3658 3659 static inline void __netif_tx_unlock_bh(struct netdev_queue *txq) 3660 { 3661 txq->xmit_lock_owner = -1; 3662 spin_unlock_bh(&txq->_xmit_lock); 3663 } 3664 3665 static inline void txq_trans_update(struct netdev_queue *txq) 3666 { 3667 if (txq->xmit_lock_owner != -1) 3668 txq->trans_start = jiffies; 3669 } 3670 3671 /* legacy drivers only, netdev_start_xmit() sets txq->trans_start */ 3672 static inline void netif_trans_update(struct net_device *dev) 3673 { 3674 struct netdev_queue *txq = netdev_get_tx_queue(dev, 0); 3675 3676 if (txq->trans_start != jiffies) 3677 txq->trans_start = jiffies; 3678 } 3679 3680 /** 3681 * netif_tx_lock - grab network device transmit lock 3682 * @dev: network device 3683 * 3684 * Get network device transmit lock 3685 */ 3686 static inline void netif_tx_lock(struct net_device *dev) 3687 { 3688 unsigned int i; 3689 int cpu; 3690 3691 spin_lock(&dev->tx_global_lock); 3692 cpu = smp_processor_id(); 3693 for (i = 0; i < dev->num_tx_queues; i++) { 3694 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 3695 3696 /* We are the only thread of execution doing a 3697 * freeze, but we have to grab the _xmit_lock in 3698 * order to synchronize with threads which are in 3699 * the ->hard_start_xmit() handler and already 3700 * checked the frozen bit. 3701 */ 3702 __netif_tx_lock(txq, cpu); 3703 set_bit(__QUEUE_STATE_FROZEN, &txq->state); 3704 __netif_tx_unlock(txq); 3705 } 3706 } 3707 3708 static inline void netif_tx_lock_bh(struct net_device *dev) 3709 { 3710 local_bh_disable(); 3711 netif_tx_lock(dev); 3712 } 3713 3714 static inline void netif_tx_unlock(struct net_device *dev) 3715 { 3716 unsigned int i; 3717 3718 for (i = 0; i < dev->num_tx_queues; i++) { 3719 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 3720 3721 /* No need to grab the _xmit_lock here. If the 3722 * queue is not stopped for another reason, we 3723 * force a schedule. 3724 */ 3725 clear_bit(__QUEUE_STATE_FROZEN, &txq->state); 3726 netif_schedule_queue(txq); 3727 } 3728 spin_unlock(&dev->tx_global_lock); 3729 } 3730 3731 static inline void netif_tx_unlock_bh(struct net_device *dev) 3732 { 3733 netif_tx_unlock(dev); 3734 local_bh_enable(); 3735 } 3736 3737 #define HARD_TX_LOCK(dev, txq, cpu) { \ 3738 if ((dev->features & NETIF_F_LLTX) == 0) { \ 3739 __netif_tx_lock(txq, cpu); \ 3740 } else { \ 3741 __netif_tx_acquire(txq); \ 3742 } \ 3743 } 3744 3745 #define HARD_TX_TRYLOCK(dev, txq) \ 3746 (((dev->features & NETIF_F_LLTX) == 0) ? \ 3747 __netif_tx_trylock(txq) : \ 3748 __netif_tx_acquire(txq)) 3749 3750 #define HARD_TX_UNLOCK(dev, txq) { \ 3751 if ((dev->features & NETIF_F_LLTX) == 0) { \ 3752 __netif_tx_unlock(txq); \ 3753 } else { \ 3754 __netif_tx_release(txq); \ 3755 } \ 3756 } 3757 3758 static inline void netif_tx_disable(struct net_device *dev) 3759 { 3760 unsigned int i; 3761 int cpu; 3762 3763 local_bh_disable(); 3764 cpu = smp_processor_id(); 3765 for (i = 0; i < dev->num_tx_queues; i++) { 3766 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 3767 3768 __netif_tx_lock(txq, cpu); 3769 netif_tx_stop_queue(txq); 3770 __netif_tx_unlock(txq); 3771 } 3772 local_bh_enable(); 3773 } 3774 3775 static inline void netif_addr_lock(struct net_device *dev) 3776 { 3777 spin_lock(&dev->addr_list_lock); 3778 } 3779 3780 static inline void netif_addr_lock_nested(struct net_device *dev) 3781 { 3782 int subclass = SINGLE_DEPTH_NESTING; 3783 3784 if (dev->netdev_ops->ndo_get_lock_subclass) 3785 subclass = dev->netdev_ops->ndo_get_lock_subclass(dev); 3786 3787 spin_lock_nested(&dev->addr_list_lock, subclass); 3788 } 3789 3790 static inline void netif_addr_lock_bh(struct net_device *dev) 3791 { 3792 spin_lock_bh(&dev->addr_list_lock); 3793 } 3794 3795 static inline void netif_addr_unlock(struct net_device *dev) 3796 { 3797 spin_unlock(&dev->addr_list_lock); 3798 } 3799 3800 static inline void netif_addr_unlock_bh(struct net_device *dev) 3801 { 3802 spin_unlock_bh(&dev->addr_list_lock); 3803 } 3804 3805 /* 3806 * dev_addrs walker. Should be used only for read access. Call with 3807 * rcu_read_lock held. 3808 */ 3809 #define for_each_dev_addr(dev, ha) \ 3810 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list) 3811 3812 /* These functions live elsewhere (drivers/net/net_init.c, but related) */ 3813 3814 void ether_setup(struct net_device *dev); 3815 3816 /* Support for loadable net-drivers */ 3817 struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name, 3818 unsigned char name_assign_type, 3819 void (*setup)(struct net_device *), 3820 unsigned int txqs, unsigned int rxqs); 3821 int dev_get_valid_name(struct net *net, struct net_device *dev, 3822 const char *name); 3823 3824 #define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \ 3825 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1) 3826 3827 #define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \ 3828 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \ 3829 count) 3830 3831 int register_netdev(struct net_device *dev); 3832 void unregister_netdev(struct net_device *dev); 3833 3834 /* General hardware address lists handling functions */ 3835 int __hw_addr_sync(struct netdev_hw_addr_list *to_list, 3836 struct netdev_hw_addr_list *from_list, int addr_len); 3837 void __hw_addr_unsync(struct netdev_hw_addr_list *to_list, 3838 struct netdev_hw_addr_list *from_list, int addr_len); 3839 int __hw_addr_sync_dev(struct netdev_hw_addr_list *list, 3840 struct net_device *dev, 3841 int (*sync)(struct net_device *, const unsigned char *), 3842 int (*unsync)(struct net_device *, 3843 const unsigned char *)); 3844 void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list, 3845 struct net_device *dev, 3846 int (*unsync)(struct net_device *, 3847 const unsigned char *)); 3848 void __hw_addr_init(struct netdev_hw_addr_list *list); 3849 3850 /* Functions used for device addresses handling */ 3851 int dev_addr_add(struct net_device *dev, const unsigned char *addr, 3852 unsigned char addr_type); 3853 int dev_addr_del(struct net_device *dev, const unsigned char *addr, 3854 unsigned char addr_type); 3855 void dev_addr_flush(struct net_device *dev); 3856 int dev_addr_init(struct net_device *dev); 3857 3858 /* Functions used for unicast addresses handling */ 3859 int dev_uc_add(struct net_device *dev, const unsigned char *addr); 3860 int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr); 3861 int dev_uc_del(struct net_device *dev, const unsigned char *addr); 3862 int dev_uc_sync(struct net_device *to, struct net_device *from); 3863 int dev_uc_sync_multiple(struct net_device *to, struct net_device *from); 3864 void dev_uc_unsync(struct net_device *to, struct net_device *from); 3865 void dev_uc_flush(struct net_device *dev); 3866 void dev_uc_init(struct net_device *dev); 3867 3868 /** 3869 * __dev_uc_sync - Synchonize device's unicast list 3870 * @dev: device to sync 3871 * @sync: function to call if address should be added 3872 * @unsync: function to call if address should be removed 3873 * 3874 * Add newly added addresses to the interface, and release 3875 * addresses that have been deleted. 3876 */ 3877 static inline int __dev_uc_sync(struct net_device *dev, 3878 int (*sync)(struct net_device *, 3879 const unsigned char *), 3880 int (*unsync)(struct net_device *, 3881 const unsigned char *)) 3882 { 3883 return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync); 3884 } 3885 3886 /** 3887 * __dev_uc_unsync - Remove synchronized addresses from device 3888 * @dev: device to sync 3889 * @unsync: function to call if address should be removed 3890 * 3891 * Remove all addresses that were added to the device by dev_uc_sync(). 3892 */ 3893 static inline void __dev_uc_unsync(struct net_device *dev, 3894 int (*unsync)(struct net_device *, 3895 const unsigned char *)) 3896 { 3897 __hw_addr_unsync_dev(&dev->uc, dev, unsync); 3898 } 3899 3900 /* Functions used for multicast addresses handling */ 3901 int dev_mc_add(struct net_device *dev, const unsigned char *addr); 3902 int dev_mc_add_global(struct net_device *dev, const unsigned char *addr); 3903 int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr); 3904 int dev_mc_del(struct net_device *dev, const unsigned char *addr); 3905 int dev_mc_del_global(struct net_device *dev, const unsigned char *addr); 3906 int dev_mc_sync(struct net_device *to, struct net_device *from); 3907 int dev_mc_sync_multiple(struct net_device *to, struct net_device *from); 3908 void dev_mc_unsync(struct net_device *to, struct net_device *from); 3909 void dev_mc_flush(struct net_device *dev); 3910 void dev_mc_init(struct net_device *dev); 3911 3912 /** 3913 * __dev_mc_sync - Synchonize device's multicast list 3914 * @dev: device to sync 3915 * @sync: function to call if address should be added 3916 * @unsync: function to call if address should be removed 3917 * 3918 * Add newly added addresses to the interface, and release 3919 * addresses that have been deleted. 3920 */ 3921 static inline int __dev_mc_sync(struct net_device *dev, 3922 int (*sync)(struct net_device *, 3923 const unsigned char *), 3924 int (*unsync)(struct net_device *, 3925 const unsigned char *)) 3926 { 3927 return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync); 3928 } 3929 3930 /** 3931 * __dev_mc_unsync - Remove synchronized addresses from device 3932 * @dev: device to sync 3933 * @unsync: function to call if address should be removed 3934 * 3935 * Remove all addresses that were added to the device by dev_mc_sync(). 3936 */ 3937 static inline void __dev_mc_unsync(struct net_device *dev, 3938 int (*unsync)(struct net_device *, 3939 const unsigned char *)) 3940 { 3941 __hw_addr_unsync_dev(&dev->mc, dev, unsync); 3942 } 3943 3944 /* Functions used for secondary unicast and multicast support */ 3945 void dev_set_rx_mode(struct net_device *dev); 3946 void __dev_set_rx_mode(struct net_device *dev); 3947 int dev_set_promiscuity(struct net_device *dev, int inc); 3948 int dev_set_allmulti(struct net_device *dev, int inc); 3949 void netdev_state_change(struct net_device *dev); 3950 void netdev_notify_peers(struct net_device *dev); 3951 void netdev_features_change(struct net_device *dev); 3952 /* Load a device via the kmod */ 3953 void dev_load(struct net *net, const char *name); 3954 struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev, 3955 struct rtnl_link_stats64 *storage); 3956 void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64, 3957 const struct net_device_stats *netdev_stats); 3958 3959 extern int netdev_max_backlog; 3960 extern int netdev_tstamp_prequeue; 3961 extern int weight_p; 3962 extern int dev_weight_rx_bias; 3963 extern int dev_weight_tx_bias; 3964 extern int dev_rx_weight; 3965 extern int dev_tx_weight; 3966 3967 bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev); 3968 struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev, 3969 struct list_head **iter); 3970 struct net_device *netdev_all_upper_get_next_dev_rcu(struct net_device *dev, 3971 struct list_head **iter); 3972 3973 /* iterate through upper list, must be called under RCU read lock */ 3974 #define netdev_for_each_upper_dev_rcu(dev, updev, iter) \ 3975 for (iter = &(dev)->adj_list.upper, \ 3976 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \ 3977 updev; \ 3978 updev = netdev_upper_get_next_dev_rcu(dev, &(iter))) 3979 3980 int netdev_walk_all_upper_dev_rcu(struct net_device *dev, 3981 int (*fn)(struct net_device *upper_dev, 3982 void *data), 3983 void *data); 3984 3985 bool netdev_has_upper_dev_all_rcu(struct net_device *dev, 3986 struct net_device *upper_dev); 3987 3988 bool netdev_has_any_upper_dev(struct net_device *dev); 3989 3990 void *netdev_lower_get_next_private(struct net_device *dev, 3991 struct list_head **iter); 3992 void *netdev_lower_get_next_private_rcu(struct net_device *dev, 3993 struct list_head **iter); 3994 3995 #define netdev_for_each_lower_private(dev, priv, iter) \ 3996 for (iter = (dev)->adj_list.lower.next, \ 3997 priv = netdev_lower_get_next_private(dev, &(iter)); \ 3998 priv; \ 3999 priv = netdev_lower_get_next_private(dev, &(iter))) 4000 4001 #define netdev_for_each_lower_private_rcu(dev, priv, iter) \ 4002 for (iter = &(dev)->adj_list.lower, \ 4003 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \ 4004 priv; \ 4005 priv = netdev_lower_get_next_private_rcu(dev, &(iter))) 4006 4007 void *netdev_lower_get_next(struct net_device *dev, 4008 struct list_head **iter); 4009 4010 #define netdev_for_each_lower_dev(dev, ldev, iter) \ 4011 for (iter = (dev)->adj_list.lower.next, \ 4012 ldev = netdev_lower_get_next(dev, &(iter)); \ 4013 ldev; \ 4014 ldev = netdev_lower_get_next(dev, &(iter))) 4015 4016 struct net_device *netdev_all_lower_get_next(struct net_device *dev, 4017 struct list_head **iter); 4018 struct net_device *netdev_all_lower_get_next_rcu(struct net_device *dev, 4019 struct list_head **iter); 4020 4021 int netdev_walk_all_lower_dev(struct net_device *dev, 4022 int (*fn)(struct net_device *lower_dev, 4023 void *data), 4024 void *data); 4025 int netdev_walk_all_lower_dev_rcu(struct net_device *dev, 4026 int (*fn)(struct net_device *lower_dev, 4027 void *data), 4028 void *data); 4029 4030 void *netdev_adjacent_get_private(struct list_head *adj_list); 4031 void *netdev_lower_get_first_private_rcu(struct net_device *dev); 4032 struct net_device *netdev_master_upper_dev_get(struct net_device *dev); 4033 struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev); 4034 int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev, 4035 struct netlink_ext_ack *extack); 4036 int netdev_master_upper_dev_link(struct net_device *dev, 4037 struct net_device *upper_dev, 4038 void *upper_priv, void *upper_info, 4039 struct netlink_ext_ack *extack); 4040 void netdev_upper_dev_unlink(struct net_device *dev, 4041 struct net_device *upper_dev); 4042 void netdev_adjacent_rename_links(struct net_device *dev, char *oldname); 4043 void *netdev_lower_dev_get_private(struct net_device *dev, 4044 struct net_device *lower_dev); 4045 void netdev_lower_state_changed(struct net_device *lower_dev, 4046 void *lower_state_info); 4047 4048 /* RSS keys are 40 or 52 bytes long */ 4049 #define NETDEV_RSS_KEY_LEN 52 4050 extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN] __read_mostly; 4051 void netdev_rss_key_fill(void *buffer, size_t len); 4052 4053 int dev_get_nest_level(struct net_device *dev); 4054 int skb_checksum_help(struct sk_buff *skb); 4055 int skb_crc32c_csum_help(struct sk_buff *skb); 4056 int skb_csum_hwoffload_help(struct sk_buff *skb, 4057 const netdev_features_t features); 4058 4059 struct sk_buff *__skb_gso_segment(struct sk_buff *skb, 4060 netdev_features_t features, bool tx_path); 4061 struct sk_buff *skb_mac_gso_segment(struct sk_buff *skb, 4062 netdev_features_t features); 4063 4064 struct netdev_bonding_info { 4065 ifslave slave; 4066 ifbond master; 4067 }; 4068 4069 struct netdev_notifier_bonding_info { 4070 struct netdev_notifier_info info; /* must be first */ 4071 struct netdev_bonding_info bonding_info; 4072 }; 4073 4074 void netdev_bonding_info_change(struct net_device *dev, 4075 struct netdev_bonding_info *bonding_info); 4076 4077 static inline 4078 struct sk_buff *skb_gso_segment(struct sk_buff *skb, netdev_features_t features) 4079 { 4080 return __skb_gso_segment(skb, features, true); 4081 } 4082 __be16 skb_network_protocol(struct sk_buff *skb, int *depth); 4083 4084 static inline bool can_checksum_protocol(netdev_features_t features, 4085 __be16 protocol) 4086 { 4087 if (protocol == htons(ETH_P_FCOE)) 4088 return !!(features & NETIF_F_FCOE_CRC); 4089 4090 /* Assume this is an IP checksum (not SCTP CRC) */ 4091 4092 if (features & NETIF_F_HW_CSUM) { 4093 /* Can checksum everything */ 4094 return true; 4095 } 4096 4097 switch (protocol) { 4098 case htons(ETH_P_IP): 4099 return !!(features & NETIF_F_IP_CSUM); 4100 case htons(ETH_P_IPV6): 4101 return !!(features & NETIF_F_IPV6_CSUM); 4102 default: 4103 return false; 4104 } 4105 } 4106 4107 #ifdef CONFIG_BUG 4108 void netdev_rx_csum_fault(struct net_device *dev); 4109 #else 4110 static inline void netdev_rx_csum_fault(struct net_device *dev) 4111 { 4112 } 4113 #endif 4114 /* rx skb timestamps */ 4115 void net_enable_timestamp(void); 4116 void net_disable_timestamp(void); 4117 4118 #ifdef CONFIG_PROC_FS 4119 int __init dev_proc_init(void); 4120 #else 4121 #define dev_proc_init() 0 4122 #endif 4123 4124 static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops, 4125 struct sk_buff *skb, struct net_device *dev, 4126 bool more) 4127 { 4128 skb->xmit_more = more ? 1 : 0; 4129 return ops->ndo_start_xmit(skb, dev); 4130 } 4131 4132 static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev, 4133 struct netdev_queue *txq, bool more) 4134 { 4135 const struct net_device_ops *ops = dev->netdev_ops; 4136 int rc; 4137 4138 rc = __netdev_start_xmit(ops, skb, dev, more); 4139 if (rc == NETDEV_TX_OK) 4140 txq_trans_update(txq); 4141 4142 return rc; 4143 } 4144 4145 int netdev_class_create_file_ns(const struct class_attribute *class_attr, 4146 const void *ns); 4147 void netdev_class_remove_file_ns(const struct class_attribute *class_attr, 4148 const void *ns); 4149 4150 static inline int netdev_class_create_file(const struct class_attribute *class_attr) 4151 { 4152 return netdev_class_create_file_ns(class_attr, NULL); 4153 } 4154 4155 static inline void netdev_class_remove_file(const struct class_attribute *class_attr) 4156 { 4157 netdev_class_remove_file_ns(class_attr, NULL); 4158 } 4159 4160 extern const struct kobj_ns_type_operations net_ns_type_operations; 4161 4162 const char *netdev_drivername(const struct net_device *dev); 4163 4164 void linkwatch_run_queue(void); 4165 4166 static inline netdev_features_t netdev_intersect_features(netdev_features_t f1, 4167 netdev_features_t f2) 4168 { 4169 if ((f1 ^ f2) & NETIF_F_HW_CSUM) { 4170 if (f1 & NETIF_F_HW_CSUM) 4171 f1 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM); 4172 else 4173 f2 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM); 4174 } 4175 4176 return f1 & f2; 4177 } 4178 4179 static inline netdev_features_t netdev_get_wanted_features( 4180 struct net_device *dev) 4181 { 4182 return (dev->features & ~dev->hw_features) | dev->wanted_features; 4183 } 4184 netdev_features_t netdev_increment_features(netdev_features_t all, 4185 netdev_features_t one, netdev_features_t mask); 4186 4187 /* Allow TSO being used on stacked device : 4188 * Performing the GSO segmentation before last device 4189 * is a performance improvement. 4190 */ 4191 static inline netdev_features_t netdev_add_tso_features(netdev_features_t features, 4192 netdev_features_t mask) 4193 { 4194 return netdev_increment_features(features, NETIF_F_ALL_TSO, mask); 4195 } 4196 4197 int __netdev_update_features(struct net_device *dev); 4198 void netdev_update_features(struct net_device *dev); 4199 void netdev_change_features(struct net_device *dev); 4200 4201 void netif_stacked_transfer_operstate(const struct net_device *rootdev, 4202 struct net_device *dev); 4203 4204 netdev_features_t passthru_features_check(struct sk_buff *skb, 4205 struct net_device *dev, 4206 netdev_features_t features); 4207 netdev_features_t netif_skb_features(struct sk_buff *skb); 4208 4209 static inline bool net_gso_ok(netdev_features_t features, int gso_type) 4210 { 4211 netdev_features_t feature = (netdev_features_t)gso_type << NETIF_F_GSO_SHIFT; 4212 4213 /* check flags correspondence */ 4214 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT)); 4215 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT)); 4216 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT)); 4217 BUILD_BUG_ON(SKB_GSO_TCP_FIXEDID != (NETIF_F_TSO_MANGLEID >> NETIF_F_GSO_SHIFT)); 4218 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT)); 4219 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT)); 4220 BUILD_BUG_ON(SKB_GSO_GRE != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT)); 4221 BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT)); 4222 BUILD_BUG_ON(SKB_GSO_IPXIP4 != (NETIF_F_GSO_IPXIP4 >> NETIF_F_GSO_SHIFT)); 4223 BUILD_BUG_ON(SKB_GSO_IPXIP6 != (NETIF_F_GSO_IPXIP6 >> NETIF_F_GSO_SHIFT)); 4224 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT)); 4225 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT)); 4226 BUILD_BUG_ON(SKB_GSO_PARTIAL != (NETIF_F_GSO_PARTIAL >> NETIF_F_GSO_SHIFT)); 4227 BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT)); 4228 BUILD_BUG_ON(SKB_GSO_SCTP != (NETIF_F_GSO_SCTP >> NETIF_F_GSO_SHIFT)); 4229 BUILD_BUG_ON(SKB_GSO_ESP != (NETIF_F_GSO_ESP >> NETIF_F_GSO_SHIFT)); 4230 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_GSO_UDP >> NETIF_F_GSO_SHIFT)); 4231 BUILD_BUG_ON(SKB_GSO_UDP_L4 != (NETIF_F_GSO_UDP_L4 >> NETIF_F_GSO_SHIFT)); 4232 4233 return (features & feature) == feature; 4234 } 4235 4236 static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features) 4237 { 4238 return net_gso_ok(features, skb_shinfo(skb)->gso_type) && 4239 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST)); 4240 } 4241 4242 static inline bool netif_needs_gso(struct sk_buff *skb, 4243 netdev_features_t features) 4244 { 4245 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) || 4246 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) && 4247 (skb->ip_summed != CHECKSUM_UNNECESSARY))); 4248 } 4249 4250 static inline void netif_set_gso_max_size(struct net_device *dev, 4251 unsigned int size) 4252 { 4253 dev->gso_max_size = size; 4254 } 4255 4256 static inline void skb_gso_error_unwind(struct sk_buff *skb, __be16 protocol, 4257 int pulled_hlen, u16 mac_offset, 4258 int mac_len) 4259 { 4260 skb->protocol = protocol; 4261 skb->encapsulation = 1; 4262 skb_push(skb, pulled_hlen); 4263 skb_reset_transport_header(skb); 4264 skb->mac_header = mac_offset; 4265 skb->network_header = skb->mac_header + mac_len; 4266 skb->mac_len = mac_len; 4267 } 4268 4269 static inline bool netif_is_macsec(const struct net_device *dev) 4270 { 4271 return dev->priv_flags & IFF_MACSEC; 4272 } 4273 4274 static inline bool netif_is_macvlan(const struct net_device *dev) 4275 { 4276 return dev->priv_flags & IFF_MACVLAN; 4277 } 4278 4279 static inline bool netif_is_macvlan_port(const struct net_device *dev) 4280 { 4281 return dev->priv_flags & IFF_MACVLAN_PORT; 4282 } 4283 4284 static inline bool netif_is_bond_master(const struct net_device *dev) 4285 { 4286 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING; 4287 } 4288 4289 static inline bool netif_is_bond_slave(const struct net_device *dev) 4290 { 4291 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING; 4292 } 4293 4294 static inline bool netif_supports_nofcs(struct net_device *dev) 4295 { 4296 return dev->priv_flags & IFF_SUPP_NOFCS; 4297 } 4298 4299 static inline bool netif_is_l3_master(const struct net_device *dev) 4300 { 4301 return dev->priv_flags & IFF_L3MDEV_MASTER; 4302 } 4303 4304 static inline bool netif_is_l3_slave(const struct net_device *dev) 4305 { 4306 return dev->priv_flags & IFF_L3MDEV_SLAVE; 4307 } 4308 4309 static inline bool netif_is_bridge_master(const struct net_device *dev) 4310 { 4311 return dev->priv_flags & IFF_EBRIDGE; 4312 } 4313 4314 static inline bool netif_is_bridge_port(const struct net_device *dev) 4315 { 4316 return dev->priv_flags & IFF_BRIDGE_PORT; 4317 } 4318 4319 static inline bool netif_is_ovs_master(const struct net_device *dev) 4320 { 4321 return dev->priv_flags & IFF_OPENVSWITCH; 4322 } 4323 4324 static inline bool netif_is_ovs_port(const struct net_device *dev) 4325 { 4326 return dev->priv_flags & IFF_OVS_DATAPATH; 4327 } 4328 4329 static inline bool netif_is_team_master(const struct net_device *dev) 4330 { 4331 return dev->priv_flags & IFF_TEAM; 4332 } 4333 4334 static inline bool netif_is_team_port(const struct net_device *dev) 4335 { 4336 return dev->priv_flags & IFF_TEAM_PORT; 4337 } 4338 4339 static inline bool netif_is_lag_master(const struct net_device *dev) 4340 { 4341 return netif_is_bond_master(dev) || netif_is_team_master(dev); 4342 } 4343 4344 static inline bool netif_is_lag_port(const struct net_device *dev) 4345 { 4346 return netif_is_bond_slave(dev) || netif_is_team_port(dev); 4347 } 4348 4349 static inline bool netif_is_rxfh_configured(const struct net_device *dev) 4350 { 4351 return dev->priv_flags & IFF_RXFH_CONFIGURED; 4352 } 4353 4354 static inline bool netif_is_failover(const struct net_device *dev) 4355 { 4356 return dev->priv_flags & IFF_FAILOVER; 4357 } 4358 4359 static inline bool netif_is_failover_slave(const struct net_device *dev) 4360 { 4361 return dev->priv_flags & IFF_FAILOVER_SLAVE; 4362 } 4363 4364 /* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */ 4365 static inline void netif_keep_dst(struct net_device *dev) 4366 { 4367 dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM); 4368 } 4369 4370 /* return true if dev can't cope with mtu frames that need vlan tag insertion */ 4371 static inline bool netif_reduces_vlan_mtu(struct net_device *dev) 4372 { 4373 /* TODO: reserve and use an additional IFF bit, if we get more users */ 4374 return dev->priv_flags & IFF_MACSEC; 4375 } 4376 4377 extern struct pernet_operations __net_initdata loopback_net_ops; 4378 4379 /* Logging, debugging and troubleshooting/diagnostic helpers. */ 4380 4381 /* netdev_printk helpers, similar to dev_printk */ 4382 4383 static inline const char *netdev_name(const struct net_device *dev) 4384 { 4385 if (!dev->name[0] || strchr(dev->name, '%')) 4386 return "(unnamed net_device)"; 4387 return dev->name; 4388 } 4389 4390 static inline bool netdev_unregistering(const struct net_device *dev) 4391 { 4392 return dev->reg_state == NETREG_UNREGISTERING; 4393 } 4394 4395 static inline const char *netdev_reg_state(const struct net_device *dev) 4396 { 4397 switch (dev->reg_state) { 4398 case NETREG_UNINITIALIZED: return " (uninitialized)"; 4399 case NETREG_REGISTERED: return ""; 4400 case NETREG_UNREGISTERING: return " (unregistering)"; 4401 case NETREG_UNREGISTERED: return " (unregistered)"; 4402 case NETREG_RELEASED: return " (released)"; 4403 case NETREG_DUMMY: return " (dummy)"; 4404 } 4405 4406 WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, dev->reg_state); 4407 return " (unknown)"; 4408 } 4409 4410 __printf(3, 4) 4411 void netdev_printk(const char *level, const struct net_device *dev, 4412 const char *format, ...); 4413 __printf(2, 3) 4414 void netdev_emerg(const struct net_device *dev, const char *format, ...); 4415 __printf(2, 3) 4416 void netdev_alert(const struct net_device *dev, const char *format, ...); 4417 __printf(2, 3) 4418 void netdev_crit(const struct net_device *dev, const char *format, ...); 4419 __printf(2, 3) 4420 void netdev_err(const struct net_device *dev, const char *format, ...); 4421 __printf(2, 3) 4422 void netdev_warn(const struct net_device *dev, const char *format, ...); 4423 __printf(2, 3) 4424 void netdev_notice(const struct net_device *dev, const char *format, ...); 4425 __printf(2, 3) 4426 void netdev_info(const struct net_device *dev, const char *format, ...); 4427 4428 #define netdev_level_once(level, dev, fmt, ...) \ 4429 do { \ 4430 static bool __print_once __read_mostly; \ 4431 \ 4432 if (!__print_once) { \ 4433 __print_once = true; \ 4434 netdev_printk(level, dev, fmt, ##__VA_ARGS__); \ 4435 } \ 4436 } while (0) 4437 4438 #define netdev_emerg_once(dev, fmt, ...) \ 4439 netdev_level_once(KERN_EMERG, dev, fmt, ##__VA_ARGS__) 4440 #define netdev_alert_once(dev, fmt, ...) \ 4441 netdev_level_once(KERN_ALERT, dev, fmt, ##__VA_ARGS__) 4442 #define netdev_crit_once(dev, fmt, ...) \ 4443 netdev_level_once(KERN_CRIT, dev, fmt, ##__VA_ARGS__) 4444 #define netdev_err_once(dev, fmt, ...) \ 4445 netdev_level_once(KERN_ERR, dev, fmt, ##__VA_ARGS__) 4446 #define netdev_warn_once(dev, fmt, ...) \ 4447 netdev_level_once(KERN_WARNING, dev, fmt, ##__VA_ARGS__) 4448 #define netdev_notice_once(dev, fmt, ...) \ 4449 netdev_level_once(KERN_NOTICE, dev, fmt, ##__VA_ARGS__) 4450 #define netdev_info_once(dev, fmt, ...) \ 4451 netdev_level_once(KERN_INFO, dev, fmt, ##__VA_ARGS__) 4452 4453 #define MODULE_ALIAS_NETDEV(device) \ 4454 MODULE_ALIAS("netdev-" device) 4455 4456 #if defined(CONFIG_DYNAMIC_DEBUG) 4457 #define netdev_dbg(__dev, format, args...) \ 4458 do { \ 4459 dynamic_netdev_dbg(__dev, format, ##args); \ 4460 } while (0) 4461 #elif defined(DEBUG) 4462 #define netdev_dbg(__dev, format, args...) \ 4463 netdev_printk(KERN_DEBUG, __dev, format, ##args) 4464 #else 4465 #define netdev_dbg(__dev, format, args...) \ 4466 ({ \ 4467 if (0) \ 4468 netdev_printk(KERN_DEBUG, __dev, format, ##args); \ 4469 }) 4470 #endif 4471 4472 #if defined(VERBOSE_DEBUG) 4473 #define netdev_vdbg netdev_dbg 4474 #else 4475 4476 #define netdev_vdbg(dev, format, args...) \ 4477 ({ \ 4478 if (0) \ 4479 netdev_printk(KERN_DEBUG, dev, format, ##args); \ 4480 0; \ 4481 }) 4482 #endif 4483 4484 /* 4485 * netdev_WARN() acts like dev_printk(), but with the key difference 4486 * of using a WARN/WARN_ON to get the message out, including the 4487 * file/line information and a backtrace. 4488 */ 4489 #define netdev_WARN(dev, format, args...) \ 4490 WARN(1, "netdevice: %s%s: " format, netdev_name(dev), \ 4491 netdev_reg_state(dev), ##args) 4492 4493 #define netdev_WARN_ONCE(dev, format, args...) \ 4494 WARN_ONCE(1, "netdevice: %s%s: " format, netdev_name(dev), \ 4495 netdev_reg_state(dev), ##args) 4496 4497 /* netif printk helpers, similar to netdev_printk */ 4498 4499 #define netif_printk(priv, type, level, dev, fmt, args...) \ 4500 do { \ 4501 if (netif_msg_##type(priv)) \ 4502 netdev_printk(level, (dev), fmt, ##args); \ 4503 } while (0) 4504 4505 #define netif_level(level, priv, type, dev, fmt, args...) \ 4506 do { \ 4507 if (netif_msg_##type(priv)) \ 4508 netdev_##level(dev, fmt, ##args); \ 4509 } while (0) 4510 4511 #define netif_emerg(priv, type, dev, fmt, args...) \ 4512 netif_level(emerg, priv, type, dev, fmt, ##args) 4513 #define netif_alert(priv, type, dev, fmt, args...) \ 4514 netif_level(alert, priv, type, dev, fmt, ##args) 4515 #define netif_crit(priv, type, dev, fmt, args...) \ 4516 netif_level(crit, priv, type, dev, fmt, ##args) 4517 #define netif_err(priv, type, dev, fmt, args...) \ 4518 netif_level(err, priv, type, dev, fmt, ##args) 4519 #define netif_warn(priv, type, dev, fmt, args...) \ 4520 netif_level(warn, priv, type, dev, fmt, ##args) 4521 #define netif_notice(priv, type, dev, fmt, args...) \ 4522 netif_level(notice, priv, type, dev, fmt, ##args) 4523 #define netif_info(priv, type, dev, fmt, args...) \ 4524 netif_level(info, priv, type, dev, fmt, ##args) 4525 4526 #if defined(CONFIG_DYNAMIC_DEBUG) 4527 #define netif_dbg(priv, type, netdev, format, args...) \ 4528 do { \ 4529 if (netif_msg_##type(priv)) \ 4530 dynamic_netdev_dbg(netdev, format, ##args); \ 4531 } while (0) 4532 #elif defined(DEBUG) 4533 #define netif_dbg(priv, type, dev, format, args...) \ 4534 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args) 4535 #else 4536 #define netif_dbg(priv, type, dev, format, args...) \ 4537 ({ \ 4538 if (0) \ 4539 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \ 4540 0; \ 4541 }) 4542 #endif 4543 4544 /* if @cond then downgrade to debug, else print at @level */ 4545 #define netif_cond_dbg(priv, type, netdev, cond, level, fmt, args...) \ 4546 do { \ 4547 if (cond) \ 4548 netif_dbg(priv, type, netdev, fmt, ##args); \ 4549 else \ 4550 netif_ ## level(priv, type, netdev, fmt, ##args); \ 4551 } while (0) 4552 4553 #if defined(VERBOSE_DEBUG) 4554 #define netif_vdbg netif_dbg 4555 #else 4556 #define netif_vdbg(priv, type, dev, format, args...) \ 4557 ({ \ 4558 if (0) \ 4559 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \ 4560 0; \ 4561 }) 4562 #endif 4563 4564 /* 4565 * The list of packet types we will receive (as opposed to discard) 4566 * and the routines to invoke. 4567 * 4568 * Why 16. Because with 16 the only overlap we get on a hash of the 4569 * low nibble of the protocol value is RARP/SNAP/X.25. 4570 * 4571 * 0800 IP 4572 * 0001 802.3 4573 * 0002 AX.25 4574 * 0004 802.2 4575 * 8035 RARP 4576 * 0005 SNAP 4577 * 0805 X.25 4578 * 0806 ARP 4579 * 8137 IPX 4580 * 0009 Localtalk 4581 * 86DD IPv6 4582 */ 4583 #define PTYPE_HASH_SIZE (16) 4584 #define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1) 4585 4586 #endif /* _LINUX_NETDEVICE_H */ 4587