1 /* 2 * INET An implementation of the TCP/IP protocol suite for the LINUX 3 * operating system. INET is implemented using the BSD Socket 4 * interface as the means of communication with the user level. 5 * 6 * Definitions for the Interfaces handler. 7 * 8 * Version: @(#)dev.h 1.0.10 08/12/93 9 * 10 * Authors: Ross Biro 11 * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG> 12 * Corey Minyard <wf-rch!minyard@relay.EU.net> 13 * Donald J. Becker, <becker@cesdis.gsfc.nasa.gov> 14 * Alan Cox, <alan@lxorguk.ukuu.org.uk> 15 * Bjorn Ekwall. <bj0rn@blox.se> 16 * Pekka Riikonen <priikone@poseidon.pspt.fi> 17 * 18 * This program is free software; you can redistribute it and/or 19 * modify it under the terms of the GNU General Public License 20 * as published by the Free Software Foundation; either version 21 * 2 of the License, or (at your option) any later version. 22 * 23 * Moved to /usr/include/linux for NET3 24 */ 25 #ifndef _LINUX_NETDEVICE_H 26 #define _LINUX_NETDEVICE_H 27 28 #include <linux/timer.h> 29 #include <linux/bug.h> 30 #include <linux/delay.h> 31 #include <linux/atomic.h> 32 #include <linux/prefetch.h> 33 #include <asm/cache.h> 34 #include <asm/byteorder.h> 35 36 #include <linux/percpu.h> 37 #include <linux/rculist.h> 38 #include <linux/workqueue.h> 39 #include <linux/dynamic_queue_limits.h> 40 41 #include <linux/ethtool.h> 42 #include <net/net_namespace.h> 43 #ifdef CONFIG_DCB 44 #include <net/dcbnl.h> 45 #endif 46 #include <net/netprio_cgroup.h> 47 #include <net/xdp.h> 48 49 #include <linux/netdev_features.h> 50 #include <linux/neighbour.h> 51 #include <uapi/linux/netdevice.h> 52 #include <uapi/linux/if_bonding.h> 53 #include <uapi/linux/pkt_cls.h> 54 #include <linux/hashtable.h> 55 56 struct netpoll_info; 57 struct device; 58 struct phy_device; 59 struct dsa_port; 60 61 /* 802.11 specific */ 62 struct wireless_dev; 63 /* 802.15.4 specific */ 64 struct wpan_dev; 65 struct mpls_dev; 66 /* UDP Tunnel offloads */ 67 struct udp_tunnel_info; 68 struct bpf_prog; 69 struct xdp_buff; 70 71 void netdev_set_default_ethtool_ops(struct net_device *dev, 72 const struct ethtool_ops *ops); 73 74 /* Backlog congestion levels */ 75 #define NET_RX_SUCCESS 0 /* keep 'em coming, baby */ 76 #define NET_RX_DROP 1 /* packet dropped */ 77 78 /* 79 * Transmit return codes: transmit return codes originate from three different 80 * namespaces: 81 * 82 * - qdisc return codes 83 * - driver transmit return codes 84 * - errno values 85 * 86 * Drivers are allowed to return any one of those in their hard_start_xmit() 87 * function. Real network devices commonly used with qdiscs should only return 88 * the driver transmit return codes though - when qdiscs are used, the actual 89 * transmission happens asynchronously, so the value is not propagated to 90 * higher layers. Virtual network devices transmit synchronously; in this case 91 * the driver transmit return codes are consumed by dev_queue_xmit(), and all 92 * others are propagated to higher layers. 93 */ 94 95 /* qdisc ->enqueue() return codes. */ 96 #define NET_XMIT_SUCCESS 0x00 97 #define NET_XMIT_DROP 0x01 /* skb dropped */ 98 #define NET_XMIT_CN 0x02 /* congestion notification */ 99 #define NET_XMIT_MASK 0x0f /* qdisc flags in net/sch_generic.h */ 100 101 /* NET_XMIT_CN is special. It does not guarantee that this packet is lost. It 102 * indicates that the device will soon be dropping packets, or already drops 103 * some packets of the same priority; prompting us to send less aggressively. */ 104 #define net_xmit_eval(e) ((e) == NET_XMIT_CN ? 0 : (e)) 105 #define net_xmit_errno(e) ((e) != NET_XMIT_CN ? -ENOBUFS : 0) 106 107 /* Driver transmit return codes */ 108 #define NETDEV_TX_MASK 0xf0 109 110 enum netdev_tx { 111 __NETDEV_TX_MIN = INT_MIN, /* make sure enum is signed */ 112 NETDEV_TX_OK = 0x00, /* driver took care of packet */ 113 NETDEV_TX_BUSY = 0x10, /* driver tx path was busy*/ 114 }; 115 typedef enum netdev_tx netdev_tx_t; 116 117 /* 118 * Current order: NETDEV_TX_MASK > NET_XMIT_MASK >= 0 is significant; 119 * hard_start_xmit() return < NET_XMIT_MASK means skb was consumed. 120 */ 121 static inline bool dev_xmit_complete(int rc) 122 { 123 /* 124 * Positive cases with an skb consumed by a driver: 125 * - successful transmission (rc == NETDEV_TX_OK) 126 * - error while transmitting (rc < 0) 127 * - error while queueing to a different device (rc & NET_XMIT_MASK) 128 */ 129 if (likely(rc < NET_XMIT_MASK)) 130 return true; 131 132 return false; 133 } 134 135 /* 136 * Compute the worst-case header length according to the protocols 137 * used. 138 */ 139 140 #if defined(CONFIG_HYPERV_NET) 141 # define LL_MAX_HEADER 128 142 #elif defined(CONFIG_WLAN) || IS_ENABLED(CONFIG_AX25) 143 # if defined(CONFIG_MAC80211_MESH) 144 # define LL_MAX_HEADER 128 145 # else 146 # define LL_MAX_HEADER 96 147 # endif 148 #else 149 # define LL_MAX_HEADER 32 150 #endif 151 152 #if !IS_ENABLED(CONFIG_NET_IPIP) && !IS_ENABLED(CONFIG_NET_IPGRE) && \ 153 !IS_ENABLED(CONFIG_IPV6_SIT) && !IS_ENABLED(CONFIG_IPV6_TUNNEL) 154 #define MAX_HEADER LL_MAX_HEADER 155 #else 156 #define MAX_HEADER (LL_MAX_HEADER + 48) 157 #endif 158 159 /* 160 * Old network device statistics. Fields are native words 161 * (unsigned long) so they can be read and written atomically. 162 */ 163 164 struct net_device_stats { 165 unsigned long rx_packets; 166 unsigned long tx_packets; 167 unsigned long rx_bytes; 168 unsigned long tx_bytes; 169 unsigned long rx_errors; 170 unsigned long tx_errors; 171 unsigned long rx_dropped; 172 unsigned long tx_dropped; 173 unsigned long multicast; 174 unsigned long collisions; 175 unsigned long rx_length_errors; 176 unsigned long rx_over_errors; 177 unsigned long rx_crc_errors; 178 unsigned long rx_frame_errors; 179 unsigned long rx_fifo_errors; 180 unsigned long rx_missed_errors; 181 unsigned long tx_aborted_errors; 182 unsigned long tx_carrier_errors; 183 unsigned long tx_fifo_errors; 184 unsigned long tx_heartbeat_errors; 185 unsigned long tx_window_errors; 186 unsigned long rx_compressed; 187 unsigned long tx_compressed; 188 }; 189 190 191 #include <linux/cache.h> 192 #include <linux/skbuff.h> 193 194 #ifdef CONFIG_RPS 195 #include <linux/static_key.h> 196 extern struct static_key rps_needed; 197 extern struct static_key rfs_needed; 198 #endif 199 200 struct neighbour; 201 struct neigh_parms; 202 struct sk_buff; 203 204 struct netdev_hw_addr { 205 struct list_head list; 206 unsigned char addr[MAX_ADDR_LEN]; 207 unsigned char type; 208 #define NETDEV_HW_ADDR_T_LAN 1 209 #define NETDEV_HW_ADDR_T_SAN 2 210 #define NETDEV_HW_ADDR_T_SLAVE 3 211 #define NETDEV_HW_ADDR_T_UNICAST 4 212 #define NETDEV_HW_ADDR_T_MULTICAST 5 213 bool global_use; 214 int sync_cnt; 215 int refcount; 216 int synced; 217 struct rcu_head rcu_head; 218 }; 219 220 struct netdev_hw_addr_list { 221 struct list_head list; 222 int count; 223 }; 224 225 #define netdev_hw_addr_list_count(l) ((l)->count) 226 #define netdev_hw_addr_list_empty(l) (netdev_hw_addr_list_count(l) == 0) 227 #define netdev_hw_addr_list_for_each(ha, l) \ 228 list_for_each_entry(ha, &(l)->list, list) 229 230 #define netdev_uc_count(dev) netdev_hw_addr_list_count(&(dev)->uc) 231 #define netdev_uc_empty(dev) netdev_hw_addr_list_empty(&(dev)->uc) 232 #define netdev_for_each_uc_addr(ha, dev) \ 233 netdev_hw_addr_list_for_each(ha, &(dev)->uc) 234 235 #define netdev_mc_count(dev) netdev_hw_addr_list_count(&(dev)->mc) 236 #define netdev_mc_empty(dev) netdev_hw_addr_list_empty(&(dev)->mc) 237 #define netdev_for_each_mc_addr(ha, dev) \ 238 netdev_hw_addr_list_for_each(ha, &(dev)->mc) 239 240 struct hh_cache { 241 unsigned int hh_len; 242 seqlock_t hh_lock; 243 244 /* cached hardware header; allow for machine alignment needs. */ 245 #define HH_DATA_MOD 16 246 #define HH_DATA_OFF(__len) \ 247 (HH_DATA_MOD - (((__len - 1) & (HH_DATA_MOD - 1)) + 1)) 248 #define HH_DATA_ALIGN(__len) \ 249 (((__len)+(HH_DATA_MOD-1))&~(HH_DATA_MOD - 1)) 250 unsigned long hh_data[HH_DATA_ALIGN(LL_MAX_HEADER) / sizeof(long)]; 251 }; 252 253 /* Reserve HH_DATA_MOD byte-aligned hard_header_len, but at least that much. 254 * Alternative is: 255 * dev->hard_header_len ? (dev->hard_header_len + 256 * (HH_DATA_MOD - 1)) & ~(HH_DATA_MOD - 1) : 0 257 * 258 * We could use other alignment values, but we must maintain the 259 * relationship HH alignment <= LL alignment. 260 */ 261 #define LL_RESERVED_SPACE(dev) \ 262 ((((dev)->hard_header_len+(dev)->needed_headroom)&~(HH_DATA_MOD - 1)) + HH_DATA_MOD) 263 #define LL_RESERVED_SPACE_EXTRA(dev,extra) \ 264 ((((dev)->hard_header_len+(dev)->needed_headroom+(extra))&~(HH_DATA_MOD - 1)) + HH_DATA_MOD) 265 266 struct header_ops { 267 int (*create) (struct sk_buff *skb, struct net_device *dev, 268 unsigned short type, const void *daddr, 269 const void *saddr, unsigned int len); 270 int (*parse)(const struct sk_buff *skb, unsigned char *haddr); 271 int (*cache)(const struct neighbour *neigh, struct hh_cache *hh, __be16 type); 272 void (*cache_update)(struct hh_cache *hh, 273 const struct net_device *dev, 274 const unsigned char *haddr); 275 bool (*validate)(const char *ll_header, unsigned int len); 276 }; 277 278 /* These flag bits are private to the generic network queueing 279 * layer; they may not be explicitly referenced by any other 280 * code. 281 */ 282 283 enum netdev_state_t { 284 __LINK_STATE_START, 285 __LINK_STATE_PRESENT, 286 __LINK_STATE_NOCARRIER, 287 __LINK_STATE_LINKWATCH_PENDING, 288 __LINK_STATE_DORMANT, 289 }; 290 291 292 /* 293 * This structure holds boot-time configured netdevice settings. They 294 * are then used in the device probing. 295 */ 296 struct netdev_boot_setup { 297 char name[IFNAMSIZ]; 298 struct ifmap map; 299 }; 300 #define NETDEV_BOOT_SETUP_MAX 8 301 302 int __init netdev_boot_setup(char *str); 303 304 /* 305 * Structure for NAPI scheduling similar to tasklet but with weighting 306 */ 307 struct napi_struct { 308 /* The poll_list must only be managed by the entity which 309 * changes the state of the NAPI_STATE_SCHED bit. This means 310 * whoever atomically sets that bit can add this napi_struct 311 * to the per-CPU poll_list, and whoever clears that bit 312 * can remove from the list right before clearing the bit. 313 */ 314 struct list_head poll_list; 315 316 unsigned long state; 317 int weight; 318 unsigned int gro_count; 319 int (*poll)(struct napi_struct *, int); 320 #ifdef CONFIG_NETPOLL 321 int poll_owner; 322 #endif 323 struct net_device *dev; 324 struct sk_buff *gro_list; 325 struct sk_buff *skb; 326 struct hrtimer timer; 327 struct list_head dev_list; 328 struct hlist_node napi_hash_node; 329 unsigned int napi_id; 330 }; 331 332 enum { 333 NAPI_STATE_SCHED, /* Poll is scheduled */ 334 NAPI_STATE_MISSED, /* reschedule a napi */ 335 NAPI_STATE_DISABLE, /* Disable pending */ 336 NAPI_STATE_NPSVC, /* Netpoll - don't dequeue from poll_list */ 337 NAPI_STATE_HASHED, /* In NAPI hash (busy polling possible) */ 338 NAPI_STATE_NO_BUSY_POLL,/* Do not add in napi_hash, no busy polling */ 339 NAPI_STATE_IN_BUSY_POLL,/* sk_busy_loop() owns this NAPI */ 340 }; 341 342 enum { 343 NAPIF_STATE_SCHED = BIT(NAPI_STATE_SCHED), 344 NAPIF_STATE_MISSED = BIT(NAPI_STATE_MISSED), 345 NAPIF_STATE_DISABLE = BIT(NAPI_STATE_DISABLE), 346 NAPIF_STATE_NPSVC = BIT(NAPI_STATE_NPSVC), 347 NAPIF_STATE_HASHED = BIT(NAPI_STATE_HASHED), 348 NAPIF_STATE_NO_BUSY_POLL = BIT(NAPI_STATE_NO_BUSY_POLL), 349 NAPIF_STATE_IN_BUSY_POLL = BIT(NAPI_STATE_IN_BUSY_POLL), 350 }; 351 352 enum gro_result { 353 GRO_MERGED, 354 GRO_MERGED_FREE, 355 GRO_HELD, 356 GRO_NORMAL, 357 GRO_DROP, 358 GRO_CONSUMED, 359 }; 360 typedef enum gro_result gro_result_t; 361 362 /* 363 * enum rx_handler_result - Possible return values for rx_handlers. 364 * @RX_HANDLER_CONSUMED: skb was consumed by rx_handler, do not process it 365 * further. 366 * @RX_HANDLER_ANOTHER: Do another round in receive path. This is indicated in 367 * case skb->dev was changed by rx_handler. 368 * @RX_HANDLER_EXACT: Force exact delivery, no wildcard. 369 * @RX_HANDLER_PASS: Do nothing, pass the skb as if no rx_handler was called. 370 * 371 * rx_handlers are functions called from inside __netif_receive_skb(), to do 372 * special processing of the skb, prior to delivery to protocol handlers. 373 * 374 * Currently, a net_device can only have a single rx_handler registered. Trying 375 * to register a second rx_handler will return -EBUSY. 376 * 377 * To register a rx_handler on a net_device, use netdev_rx_handler_register(). 378 * To unregister a rx_handler on a net_device, use 379 * netdev_rx_handler_unregister(). 380 * 381 * Upon return, rx_handler is expected to tell __netif_receive_skb() what to 382 * do with the skb. 383 * 384 * If the rx_handler consumed the skb in some way, it should return 385 * RX_HANDLER_CONSUMED. This is appropriate when the rx_handler arranged for 386 * the skb to be delivered in some other way. 387 * 388 * If the rx_handler changed skb->dev, to divert the skb to another 389 * net_device, it should return RX_HANDLER_ANOTHER. The rx_handler for the 390 * new device will be called if it exists. 391 * 392 * If the rx_handler decides the skb should be ignored, it should return 393 * RX_HANDLER_EXACT. The skb will only be delivered to protocol handlers that 394 * are registered on exact device (ptype->dev == skb->dev). 395 * 396 * If the rx_handler didn't change skb->dev, but wants the skb to be normally 397 * delivered, it should return RX_HANDLER_PASS. 398 * 399 * A device without a registered rx_handler will behave as if rx_handler 400 * returned RX_HANDLER_PASS. 401 */ 402 403 enum rx_handler_result { 404 RX_HANDLER_CONSUMED, 405 RX_HANDLER_ANOTHER, 406 RX_HANDLER_EXACT, 407 RX_HANDLER_PASS, 408 }; 409 typedef enum rx_handler_result rx_handler_result_t; 410 typedef rx_handler_result_t rx_handler_func_t(struct sk_buff **pskb); 411 412 void __napi_schedule(struct napi_struct *n); 413 void __napi_schedule_irqoff(struct napi_struct *n); 414 415 static inline bool napi_disable_pending(struct napi_struct *n) 416 { 417 return test_bit(NAPI_STATE_DISABLE, &n->state); 418 } 419 420 bool napi_schedule_prep(struct napi_struct *n); 421 422 /** 423 * napi_schedule - schedule NAPI poll 424 * @n: NAPI context 425 * 426 * Schedule NAPI poll routine to be called if it is not already 427 * running. 428 */ 429 static inline void napi_schedule(struct napi_struct *n) 430 { 431 if (napi_schedule_prep(n)) 432 __napi_schedule(n); 433 } 434 435 /** 436 * napi_schedule_irqoff - schedule NAPI poll 437 * @n: NAPI context 438 * 439 * Variant of napi_schedule(), assuming hard irqs are masked. 440 */ 441 static inline void napi_schedule_irqoff(struct napi_struct *n) 442 { 443 if (napi_schedule_prep(n)) 444 __napi_schedule_irqoff(n); 445 } 446 447 /* Try to reschedule poll. Called by dev->poll() after napi_complete(). */ 448 static inline bool napi_reschedule(struct napi_struct *napi) 449 { 450 if (napi_schedule_prep(napi)) { 451 __napi_schedule(napi); 452 return true; 453 } 454 return false; 455 } 456 457 bool napi_complete_done(struct napi_struct *n, int work_done); 458 /** 459 * napi_complete - NAPI processing complete 460 * @n: NAPI context 461 * 462 * Mark NAPI processing as complete. 463 * Consider using napi_complete_done() instead. 464 * Return false if device should avoid rearming interrupts. 465 */ 466 static inline bool napi_complete(struct napi_struct *n) 467 { 468 return napi_complete_done(n, 0); 469 } 470 471 /** 472 * napi_hash_del - remove a NAPI from global table 473 * @napi: NAPI context 474 * 475 * Warning: caller must observe RCU grace period 476 * before freeing memory containing @napi, if 477 * this function returns true. 478 * Note: core networking stack automatically calls it 479 * from netif_napi_del(). 480 * Drivers might want to call this helper to combine all 481 * the needed RCU grace periods into a single one. 482 */ 483 bool napi_hash_del(struct napi_struct *napi); 484 485 /** 486 * napi_disable - prevent NAPI from scheduling 487 * @n: NAPI context 488 * 489 * Stop NAPI from being scheduled on this context. 490 * Waits till any outstanding processing completes. 491 */ 492 void napi_disable(struct napi_struct *n); 493 494 /** 495 * napi_enable - enable NAPI scheduling 496 * @n: NAPI context 497 * 498 * Resume NAPI from being scheduled on this context. 499 * Must be paired with napi_disable. 500 */ 501 static inline void napi_enable(struct napi_struct *n) 502 { 503 BUG_ON(!test_bit(NAPI_STATE_SCHED, &n->state)); 504 smp_mb__before_atomic(); 505 clear_bit(NAPI_STATE_SCHED, &n->state); 506 clear_bit(NAPI_STATE_NPSVC, &n->state); 507 } 508 509 /** 510 * napi_synchronize - wait until NAPI is not running 511 * @n: NAPI context 512 * 513 * Wait until NAPI is done being scheduled on this context. 514 * Waits till any outstanding processing completes but 515 * does not disable future activations. 516 */ 517 static inline void napi_synchronize(const struct napi_struct *n) 518 { 519 if (IS_ENABLED(CONFIG_SMP)) 520 while (test_bit(NAPI_STATE_SCHED, &n->state)) 521 msleep(1); 522 else 523 barrier(); 524 } 525 526 enum netdev_queue_state_t { 527 __QUEUE_STATE_DRV_XOFF, 528 __QUEUE_STATE_STACK_XOFF, 529 __QUEUE_STATE_FROZEN, 530 }; 531 532 #define QUEUE_STATE_DRV_XOFF (1 << __QUEUE_STATE_DRV_XOFF) 533 #define QUEUE_STATE_STACK_XOFF (1 << __QUEUE_STATE_STACK_XOFF) 534 #define QUEUE_STATE_FROZEN (1 << __QUEUE_STATE_FROZEN) 535 536 #define QUEUE_STATE_ANY_XOFF (QUEUE_STATE_DRV_XOFF | QUEUE_STATE_STACK_XOFF) 537 #define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \ 538 QUEUE_STATE_FROZEN) 539 #define QUEUE_STATE_DRV_XOFF_OR_FROZEN (QUEUE_STATE_DRV_XOFF | \ 540 QUEUE_STATE_FROZEN) 541 542 /* 543 * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue. The 544 * netif_tx_* functions below are used to manipulate this flag. The 545 * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit 546 * queue independently. The netif_xmit_*stopped functions below are called 547 * to check if the queue has been stopped by the driver or stack (either 548 * of the XOFF bits are set in the state). Drivers should not need to call 549 * netif_xmit*stopped functions, they should only be using netif_tx_*. 550 */ 551 552 struct netdev_queue { 553 /* 554 * read-mostly part 555 */ 556 struct net_device *dev; 557 struct Qdisc __rcu *qdisc; 558 struct Qdisc *qdisc_sleeping; 559 #ifdef CONFIG_SYSFS 560 struct kobject kobj; 561 #endif 562 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA) 563 int numa_node; 564 #endif 565 unsigned long tx_maxrate; 566 /* 567 * Number of TX timeouts for this queue 568 * (/sys/class/net/DEV/Q/trans_timeout) 569 */ 570 unsigned long trans_timeout; 571 /* 572 * write-mostly part 573 */ 574 spinlock_t _xmit_lock ____cacheline_aligned_in_smp; 575 int xmit_lock_owner; 576 /* 577 * Time (in jiffies) of last Tx 578 */ 579 unsigned long trans_start; 580 581 unsigned long state; 582 583 #ifdef CONFIG_BQL 584 struct dql dql; 585 #endif 586 } ____cacheline_aligned_in_smp; 587 588 static inline int netdev_queue_numa_node_read(const struct netdev_queue *q) 589 { 590 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA) 591 return q->numa_node; 592 #else 593 return NUMA_NO_NODE; 594 #endif 595 } 596 597 static inline void netdev_queue_numa_node_write(struct netdev_queue *q, int node) 598 { 599 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA) 600 q->numa_node = node; 601 #endif 602 } 603 604 #ifdef CONFIG_RPS 605 /* 606 * This structure holds an RPS map which can be of variable length. The 607 * map is an array of CPUs. 608 */ 609 struct rps_map { 610 unsigned int len; 611 struct rcu_head rcu; 612 u16 cpus[0]; 613 }; 614 #define RPS_MAP_SIZE(_num) (sizeof(struct rps_map) + ((_num) * sizeof(u16))) 615 616 /* 617 * The rps_dev_flow structure contains the mapping of a flow to a CPU, the 618 * tail pointer for that CPU's input queue at the time of last enqueue, and 619 * a hardware filter index. 620 */ 621 struct rps_dev_flow { 622 u16 cpu; 623 u16 filter; 624 unsigned int last_qtail; 625 }; 626 #define RPS_NO_FILTER 0xffff 627 628 /* 629 * The rps_dev_flow_table structure contains a table of flow mappings. 630 */ 631 struct rps_dev_flow_table { 632 unsigned int mask; 633 struct rcu_head rcu; 634 struct rps_dev_flow flows[0]; 635 }; 636 #define RPS_DEV_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_dev_flow_table) + \ 637 ((_num) * sizeof(struct rps_dev_flow))) 638 639 /* 640 * The rps_sock_flow_table contains mappings of flows to the last CPU 641 * on which they were processed by the application (set in recvmsg). 642 * Each entry is a 32bit value. Upper part is the high-order bits 643 * of flow hash, lower part is CPU number. 644 * rps_cpu_mask is used to partition the space, depending on number of 645 * possible CPUs : rps_cpu_mask = roundup_pow_of_two(nr_cpu_ids) - 1 646 * For example, if 64 CPUs are possible, rps_cpu_mask = 0x3f, 647 * meaning we use 32-6=26 bits for the hash. 648 */ 649 struct rps_sock_flow_table { 650 u32 mask; 651 652 u32 ents[0] ____cacheline_aligned_in_smp; 653 }; 654 #define RPS_SOCK_FLOW_TABLE_SIZE(_num) (offsetof(struct rps_sock_flow_table, ents[_num])) 655 656 #define RPS_NO_CPU 0xffff 657 658 extern u32 rps_cpu_mask; 659 extern struct rps_sock_flow_table __rcu *rps_sock_flow_table; 660 661 static inline void rps_record_sock_flow(struct rps_sock_flow_table *table, 662 u32 hash) 663 { 664 if (table && hash) { 665 unsigned int index = hash & table->mask; 666 u32 val = hash & ~rps_cpu_mask; 667 668 /* We only give a hint, preemption can change CPU under us */ 669 val |= raw_smp_processor_id(); 670 671 if (table->ents[index] != val) 672 table->ents[index] = val; 673 } 674 } 675 676 #ifdef CONFIG_RFS_ACCEL 677 bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index, u32 flow_id, 678 u16 filter_id); 679 #endif 680 #endif /* CONFIG_RPS */ 681 682 /* This structure contains an instance of an RX queue. */ 683 struct netdev_rx_queue { 684 #ifdef CONFIG_RPS 685 struct rps_map __rcu *rps_map; 686 struct rps_dev_flow_table __rcu *rps_flow_table; 687 #endif 688 struct kobject kobj; 689 struct net_device *dev; 690 struct xdp_rxq_info xdp_rxq; 691 } ____cacheline_aligned_in_smp; 692 693 /* 694 * RX queue sysfs structures and functions. 695 */ 696 struct rx_queue_attribute { 697 struct attribute attr; 698 ssize_t (*show)(struct netdev_rx_queue *queue, char *buf); 699 ssize_t (*store)(struct netdev_rx_queue *queue, 700 const char *buf, size_t len); 701 }; 702 703 #ifdef CONFIG_XPS 704 /* 705 * This structure holds an XPS map which can be of variable length. The 706 * map is an array of queues. 707 */ 708 struct xps_map { 709 unsigned int len; 710 unsigned int alloc_len; 711 struct rcu_head rcu; 712 u16 queues[0]; 713 }; 714 #define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16))) 715 #define XPS_MIN_MAP_ALLOC ((L1_CACHE_ALIGN(offsetof(struct xps_map, queues[1])) \ 716 - sizeof(struct xps_map)) / sizeof(u16)) 717 718 /* 719 * This structure holds all XPS maps for device. Maps are indexed by CPU. 720 */ 721 struct xps_dev_maps { 722 struct rcu_head rcu; 723 struct xps_map __rcu *cpu_map[0]; 724 }; 725 #define XPS_DEV_MAPS_SIZE(_tcs) (sizeof(struct xps_dev_maps) + \ 726 (nr_cpu_ids * (_tcs) * sizeof(struct xps_map *))) 727 #endif /* CONFIG_XPS */ 728 729 #define TC_MAX_QUEUE 16 730 #define TC_BITMASK 15 731 /* HW offloaded queuing disciplines txq count and offset maps */ 732 struct netdev_tc_txq { 733 u16 count; 734 u16 offset; 735 }; 736 737 #if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE) 738 /* 739 * This structure is to hold information about the device 740 * configured to run FCoE protocol stack. 741 */ 742 struct netdev_fcoe_hbainfo { 743 char manufacturer[64]; 744 char serial_number[64]; 745 char hardware_version[64]; 746 char driver_version[64]; 747 char optionrom_version[64]; 748 char firmware_version[64]; 749 char model[256]; 750 char model_description[256]; 751 }; 752 #endif 753 754 #define MAX_PHYS_ITEM_ID_LEN 32 755 756 /* This structure holds a unique identifier to identify some 757 * physical item (port for example) used by a netdevice. 758 */ 759 struct netdev_phys_item_id { 760 unsigned char id[MAX_PHYS_ITEM_ID_LEN]; 761 unsigned char id_len; 762 }; 763 764 static inline bool netdev_phys_item_id_same(struct netdev_phys_item_id *a, 765 struct netdev_phys_item_id *b) 766 { 767 return a->id_len == b->id_len && 768 memcmp(a->id, b->id, a->id_len) == 0; 769 } 770 771 typedef u16 (*select_queue_fallback_t)(struct net_device *dev, 772 struct sk_buff *skb); 773 774 enum tc_setup_type { 775 TC_SETUP_QDISC_MQPRIO, 776 TC_SETUP_CLSU32, 777 TC_SETUP_CLSFLOWER, 778 TC_SETUP_CLSMATCHALL, 779 TC_SETUP_CLSBPF, 780 TC_SETUP_BLOCK, 781 TC_SETUP_QDISC_CBS, 782 TC_SETUP_QDISC_RED, 783 }; 784 785 /* These structures hold the attributes of bpf state that are being passed 786 * to the netdevice through the bpf op. 787 */ 788 enum bpf_netdev_command { 789 /* Set or clear a bpf program used in the earliest stages of packet 790 * rx. The prog will have been loaded as BPF_PROG_TYPE_XDP. The callee 791 * is responsible for calling bpf_prog_put on any old progs that are 792 * stored. In case of error, the callee need not release the new prog 793 * reference, but on success it takes ownership and must bpf_prog_put 794 * when it is no longer used. 795 */ 796 XDP_SETUP_PROG, 797 XDP_SETUP_PROG_HW, 798 /* Check if a bpf program is set on the device. The callee should 799 * set @prog_attached to one of XDP_ATTACHED_* values, note that "true" 800 * is equivalent to XDP_ATTACHED_DRV. 801 */ 802 XDP_QUERY_PROG, 803 /* BPF program for offload callbacks, invoked at program load time. */ 804 BPF_OFFLOAD_VERIFIER_PREP, 805 BPF_OFFLOAD_TRANSLATE, 806 BPF_OFFLOAD_DESTROY, 807 }; 808 809 struct bpf_prog_offload_ops; 810 struct netlink_ext_ack; 811 812 struct netdev_bpf { 813 enum bpf_netdev_command command; 814 union { 815 /* XDP_SETUP_PROG */ 816 struct { 817 u32 flags; 818 struct bpf_prog *prog; 819 struct netlink_ext_ack *extack; 820 }; 821 /* XDP_QUERY_PROG */ 822 struct { 823 u8 prog_attached; 824 u32 prog_id; 825 /* flags with which program was installed */ 826 u32 prog_flags; 827 }; 828 /* BPF_OFFLOAD_VERIFIER_PREP */ 829 struct { 830 struct bpf_prog *prog; 831 const struct bpf_prog_offload_ops *ops; /* callee set */ 832 } verifier; 833 /* BPF_OFFLOAD_TRANSLATE, BPF_OFFLOAD_DESTROY */ 834 struct { 835 struct bpf_prog *prog; 836 } offload; 837 }; 838 }; 839 840 #ifdef CONFIG_XFRM_OFFLOAD 841 struct xfrmdev_ops { 842 int (*xdo_dev_state_add) (struct xfrm_state *x); 843 void (*xdo_dev_state_delete) (struct xfrm_state *x); 844 void (*xdo_dev_state_free) (struct xfrm_state *x); 845 bool (*xdo_dev_offload_ok) (struct sk_buff *skb, 846 struct xfrm_state *x); 847 }; 848 #endif 849 850 struct dev_ifalias { 851 struct rcu_head rcuhead; 852 char ifalias[]; 853 }; 854 855 /* 856 * This structure defines the management hooks for network devices. 857 * The following hooks can be defined; unless noted otherwise, they are 858 * optional and can be filled with a null pointer. 859 * 860 * int (*ndo_init)(struct net_device *dev); 861 * This function is called once when a network device is registered. 862 * The network device can use this for any late stage initialization 863 * or semantic validation. It can fail with an error code which will 864 * be propagated back to register_netdev. 865 * 866 * void (*ndo_uninit)(struct net_device *dev); 867 * This function is called when device is unregistered or when registration 868 * fails. It is not called if init fails. 869 * 870 * int (*ndo_open)(struct net_device *dev); 871 * This function is called when a network device transitions to the up 872 * state. 873 * 874 * int (*ndo_stop)(struct net_device *dev); 875 * This function is called when a network device transitions to the down 876 * state. 877 * 878 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb, 879 * struct net_device *dev); 880 * Called when a packet needs to be transmitted. 881 * Returns NETDEV_TX_OK. Can return NETDEV_TX_BUSY, but you should stop 882 * the queue before that can happen; it's for obsolete devices and weird 883 * corner cases, but the stack really does a non-trivial amount 884 * of useless work if you return NETDEV_TX_BUSY. 885 * Required; cannot be NULL. 886 * 887 * netdev_features_t (*ndo_features_check)(struct sk_buff *skb, 888 * struct net_device *dev 889 * netdev_features_t features); 890 * Called by core transmit path to determine if device is capable of 891 * performing offload operations on a given packet. This is to give 892 * the device an opportunity to implement any restrictions that cannot 893 * be otherwise expressed by feature flags. The check is called with 894 * the set of features that the stack has calculated and it returns 895 * those the driver believes to be appropriate. 896 * 897 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb, 898 * void *accel_priv, select_queue_fallback_t fallback); 899 * Called to decide which queue to use when device supports multiple 900 * transmit queues. 901 * 902 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags); 903 * This function is called to allow device receiver to make 904 * changes to configuration when multicast or promiscuous is enabled. 905 * 906 * void (*ndo_set_rx_mode)(struct net_device *dev); 907 * This function is called device changes address list filtering. 908 * If driver handles unicast address filtering, it should set 909 * IFF_UNICAST_FLT in its priv_flags. 910 * 911 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr); 912 * This function is called when the Media Access Control address 913 * needs to be changed. If this interface is not defined, the 914 * MAC address can not be changed. 915 * 916 * int (*ndo_validate_addr)(struct net_device *dev); 917 * Test if Media Access Control address is valid for the device. 918 * 919 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd); 920 * Called when a user requests an ioctl which can't be handled by 921 * the generic interface code. If not defined ioctls return 922 * not supported error code. 923 * 924 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map); 925 * Used to set network devices bus interface parameters. This interface 926 * is retained for legacy reasons; new devices should use the bus 927 * interface (PCI) for low level management. 928 * 929 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu); 930 * Called when a user wants to change the Maximum Transfer Unit 931 * of a device. 932 * 933 * void (*ndo_tx_timeout)(struct net_device *dev); 934 * Callback used when the transmitter has not made any progress 935 * for dev->watchdog ticks. 936 * 937 * void (*ndo_get_stats64)(struct net_device *dev, 938 * struct rtnl_link_stats64 *storage); 939 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev); 940 * Called when a user wants to get the network device usage 941 * statistics. Drivers must do one of the following: 942 * 1. Define @ndo_get_stats64 to fill in a zero-initialised 943 * rtnl_link_stats64 structure passed by the caller. 944 * 2. Define @ndo_get_stats to update a net_device_stats structure 945 * (which should normally be dev->stats) and return a pointer to 946 * it. The structure may be changed asynchronously only if each 947 * field is written atomically. 948 * 3. Update dev->stats asynchronously and atomically, and define 949 * neither operation. 950 * 951 * bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id) 952 * Return true if this device supports offload stats of this attr_id. 953 * 954 * int (*ndo_get_offload_stats)(int attr_id, const struct net_device *dev, 955 * void *attr_data) 956 * Get statistics for offload operations by attr_id. Write it into the 957 * attr_data pointer. 958 * 959 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16 vid); 960 * If device supports VLAN filtering this function is called when a 961 * VLAN id is registered. 962 * 963 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, __be16 proto, u16 vid); 964 * If device supports VLAN filtering this function is called when a 965 * VLAN id is unregistered. 966 * 967 * void (*ndo_poll_controller)(struct net_device *dev); 968 * 969 * SR-IOV management functions. 970 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac); 971 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan, 972 * u8 qos, __be16 proto); 973 * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate, 974 * int max_tx_rate); 975 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting); 976 * int (*ndo_set_vf_trust)(struct net_device *dev, int vf, bool setting); 977 * int (*ndo_get_vf_config)(struct net_device *dev, 978 * int vf, struct ifla_vf_info *ivf); 979 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state); 980 * int (*ndo_set_vf_port)(struct net_device *dev, int vf, 981 * struct nlattr *port[]); 982 * 983 * Enable or disable the VF ability to query its RSS Redirection Table and 984 * Hash Key. This is needed since on some devices VF share this information 985 * with PF and querying it may introduce a theoretical security risk. 986 * int (*ndo_set_vf_rss_query_en)(struct net_device *dev, int vf, bool setting); 987 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb); 988 * int (*ndo_setup_tc)(struct net_device *dev, enum tc_setup_type type, 989 * void *type_data); 990 * Called to setup any 'tc' scheduler, classifier or action on @dev. 991 * This is always called from the stack with the rtnl lock held and netif 992 * tx queues stopped. This allows the netdevice to perform queue 993 * management safely. 994 * 995 * Fiber Channel over Ethernet (FCoE) offload functions. 996 * int (*ndo_fcoe_enable)(struct net_device *dev); 997 * Called when the FCoE protocol stack wants to start using LLD for FCoE 998 * so the underlying device can perform whatever needed configuration or 999 * initialization to support acceleration of FCoE traffic. 1000 * 1001 * int (*ndo_fcoe_disable)(struct net_device *dev); 1002 * Called when the FCoE protocol stack wants to stop using LLD for FCoE 1003 * so the underlying device can perform whatever needed clean-ups to 1004 * stop supporting acceleration of FCoE traffic. 1005 * 1006 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid, 1007 * struct scatterlist *sgl, unsigned int sgc); 1008 * Called when the FCoE Initiator wants to initialize an I/O that 1009 * is a possible candidate for Direct Data Placement (DDP). The LLD can 1010 * perform necessary setup and returns 1 to indicate the device is set up 1011 * successfully to perform DDP on this I/O, otherwise this returns 0. 1012 * 1013 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid); 1014 * Called when the FCoE Initiator/Target is done with the DDPed I/O as 1015 * indicated by the FC exchange id 'xid', so the underlying device can 1016 * clean up and reuse resources for later DDP requests. 1017 * 1018 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid, 1019 * struct scatterlist *sgl, unsigned int sgc); 1020 * Called when the FCoE Target wants to initialize an I/O that 1021 * is a possible candidate for Direct Data Placement (DDP). The LLD can 1022 * perform necessary setup and returns 1 to indicate the device is set up 1023 * successfully to perform DDP on this I/O, otherwise this returns 0. 1024 * 1025 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev, 1026 * struct netdev_fcoe_hbainfo *hbainfo); 1027 * Called when the FCoE Protocol stack wants information on the underlying 1028 * device. This information is utilized by the FCoE protocol stack to 1029 * register attributes with Fiber Channel management service as per the 1030 * FC-GS Fabric Device Management Information(FDMI) specification. 1031 * 1032 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type); 1033 * Called when the underlying device wants to override default World Wide 1034 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own 1035 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE 1036 * protocol stack to use. 1037 * 1038 * RFS acceleration. 1039 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb, 1040 * u16 rxq_index, u32 flow_id); 1041 * Set hardware filter for RFS. rxq_index is the target queue index; 1042 * flow_id is a flow ID to be passed to rps_may_expire_flow() later. 1043 * Return the filter ID on success, or a negative error code. 1044 * 1045 * Slave management functions (for bridge, bonding, etc). 1046 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev); 1047 * Called to make another netdev an underling. 1048 * 1049 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev); 1050 * Called to release previously enslaved netdev. 1051 * 1052 * Feature/offload setting functions. 1053 * netdev_features_t (*ndo_fix_features)(struct net_device *dev, 1054 * netdev_features_t features); 1055 * Adjusts the requested feature flags according to device-specific 1056 * constraints, and returns the resulting flags. Must not modify 1057 * the device state. 1058 * 1059 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features); 1060 * Called to update device configuration to new features. Passed 1061 * feature set might be less than what was returned by ndo_fix_features()). 1062 * Must return >0 or -errno if it changed dev->features itself. 1063 * 1064 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[], 1065 * struct net_device *dev, 1066 * const unsigned char *addr, u16 vid, u16 flags) 1067 * Adds an FDB entry to dev for addr. 1068 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[], 1069 * struct net_device *dev, 1070 * const unsigned char *addr, u16 vid) 1071 * Deletes the FDB entry from dev coresponding to addr. 1072 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb, 1073 * struct net_device *dev, struct net_device *filter_dev, 1074 * int *idx) 1075 * Used to add FDB entries to dump requests. Implementers should add 1076 * entries to skb and update idx with the number of entries. 1077 * 1078 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh, 1079 * u16 flags) 1080 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq, 1081 * struct net_device *dev, u32 filter_mask, 1082 * int nlflags) 1083 * int (*ndo_bridge_dellink)(struct net_device *dev, struct nlmsghdr *nlh, 1084 * u16 flags); 1085 * 1086 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier); 1087 * Called to change device carrier. Soft-devices (like dummy, team, etc) 1088 * which do not represent real hardware may define this to allow their 1089 * userspace components to manage their virtual carrier state. Devices 1090 * that determine carrier state from physical hardware properties (eg 1091 * network cables) or protocol-dependent mechanisms (eg 1092 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function. 1093 * 1094 * int (*ndo_get_phys_port_id)(struct net_device *dev, 1095 * struct netdev_phys_item_id *ppid); 1096 * Called to get ID of physical port of this device. If driver does 1097 * not implement this, it is assumed that the hw is not able to have 1098 * multiple net devices on single physical port. 1099 * 1100 * void (*ndo_udp_tunnel_add)(struct net_device *dev, 1101 * struct udp_tunnel_info *ti); 1102 * Called by UDP tunnel to notify a driver about the UDP port and socket 1103 * address family that a UDP tunnel is listnening to. It is called only 1104 * when a new port starts listening. The operation is protected by the 1105 * RTNL. 1106 * 1107 * void (*ndo_udp_tunnel_del)(struct net_device *dev, 1108 * struct udp_tunnel_info *ti); 1109 * Called by UDP tunnel to notify the driver about a UDP port and socket 1110 * address family that the UDP tunnel is not listening to anymore. The 1111 * operation is protected by the RTNL. 1112 * 1113 * void* (*ndo_dfwd_add_station)(struct net_device *pdev, 1114 * struct net_device *dev) 1115 * Called by upper layer devices to accelerate switching or other 1116 * station functionality into hardware. 'pdev is the lowerdev 1117 * to use for the offload and 'dev' is the net device that will 1118 * back the offload. Returns a pointer to the private structure 1119 * the upper layer will maintain. 1120 * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv) 1121 * Called by upper layer device to delete the station created 1122 * by 'ndo_dfwd_add_station'. 'pdev' is the net device backing 1123 * the station and priv is the structure returned by the add 1124 * operation. 1125 * int (*ndo_set_tx_maxrate)(struct net_device *dev, 1126 * int queue_index, u32 maxrate); 1127 * Called when a user wants to set a max-rate limitation of specific 1128 * TX queue. 1129 * int (*ndo_get_iflink)(const struct net_device *dev); 1130 * Called to get the iflink value of this device. 1131 * void (*ndo_change_proto_down)(struct net_device *dev, 1132 * bool proto_down); 1133 * This function is used to pass protocol port error state information 1134 * to the switch driver. The switch driver can react to the proto_down 1135 * by doing a phys down on the associated switch port. 1136 * int (*ndo_fill_metadata_dst)(struct net_device *dev, struct sk_buff *skb); 1137 * This function is used to get egress tunnel information for given skb. 1138 * This is useful for retrieving outer tunnel header parameters while 1139 * sampling packet. 1140 * void (*ndo_set_rx_headroom)(struct net_device *dev, int needed_headroom); 1141 * This function is used to specify the headroom that the skb must 1142 * consider when allocation skb during packet reception. Setting 1143 * appropriate rx headroom value allows avoiding skb head copy on 1144 * forward. Setting a negative value resets the rx headroom to the 1145 * default value. 1146 * int (*ndo_bpf)(struct net_device *dev, struct netdev_bpf *bpf); 1147 * This function is used to set or query state related to XDP on the 1148 * netdevice and manage BPF offload. See definition of 1149 * enum bpf_netdev_command for details. 1150 * int (*ndo_xdp_xmit)(struct net_device *dev, struct xdp_buff *xdp); 1151 * This function is used to submit a XDP packet for transmit on a 1152 * netdevice. 1153 * void (*ndo_xdp_flush)(struct net_device *dev); 1154 * This function is used to inform the driver to flush a particular 1155 * xdp tx queue. Must be called on same CPU as xdp_xmit. 1156 */ 1157 struct net_device_ops { 1158 int (*ndo_init)(struct net_device *dev); 1159 void (*ndo_uninit)(struct net_device *dev); 1160 int (*ndo_open)(struct net_device *dev); 1161 int (*ndo_stop)(struct net_device *dev); 1162 netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb, 1163 struct net_device *dev); 1164 netdev_features_t (*ndo_features_check)(struct sk_buff *skb, 1165 struct net_device *dev, 1166 netdev_features_t features); 1167 u16 (*ndo_select_queue)(struct net_device *dev, 1168 struct sk_buff *skb, 1169 void *accel_priv, 1170 select_queue_fallback_t fallback); 1171 void (*ndo_change_rx_flags)(struct net_device *dev, 1172 int flags); 1173 void (*ndo_set_rx_mode)(struct net_device *dev); 1174 int (*ndo_set_mac_address)(struct net_device *dev, 1175 void *addr); 1176 int (*ndo_validate_addr)(struct net_device *dev); 1177 int (*ndo_do_ioctl)(struct net_device *dev, 1178 struct ifreq *ifr, int cmd); 1179 int (*ndo_set_config)(struct net_device *dev, 1180 struct ifmap *map); 1181 int (*ndo_change_mtu)(struct net_device *dev, 1182 int new_mtu); 1183 int (*ndo_neigh_setup)(struct net_device *dev, 1184 struct neigh_parms *); 1185 void (*ndo_tx_timeout) (struct net_device *dev); 1186 1187 void (*ndo_get_stats64)(struct net_device *dev, 1188 struct rtnl_link_stats64 *storage); 1189 bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id); 1190 int (*ndo_get_offload_stats)(int attr_id, 1191 const struct net_device *dev, 1192 void *attr_data); 1193 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev); 1194 1195 int (*ndo_vlan_rx_add_vid)(struct net_device *dev, 1196 __be16 proto, u16 vid); 1197 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, 1198 __be16 proto, u16 vid); 1199 #ifdef CONFIG_NET_POLL_CONTROLLER 1200 void (*ndo_poll_controller)(struct net_device *dev); 1201 int (*ndo_netpoll_setup)(struct net_device *dev, 1202 struct netpoll_info *info); 1203 void (*ndo_netpoll_cleanup)(struct net_device *dev); 1204 #endif 1205 int (*ndo_set_vf_mac)(struct net_device *dev, 1206 int queue, u8 *mac); 1207 int (*ndo_set_vf_vlan)(struct net_device *dev, 1208 int queue, u16 vlan, 1209 u8 qos, __be16 proto); 1210 int (*ndo_set_vf_rate)(struct net_device *dev, 1211 int vf, int min_tx_rate, 1212 int max_tx_rate); 1213 int (*ndo_set_vf_spoofchk)(struct net_device *dev, 1214 int vf, bool setting); 1215 int (*ndo_set_vf_trust)(struct net_device *dev, 1216 int vf, bool setting); 1217 int (*ndo_get_vf_config)(struct net_device *dev, 1218 int vf, 1219 struct ifla_vf_info *ivf); 1220 int (*ndo_set_vf_link_state)(struct net_device *dev, 1221 int vf, int link_state); 1222 int (*ndo_get_vf_stats)(struct net_device *dev, 1223 int vf, 1224 struct ifla_vf_stats 1225 *vf_stats); 1226 int (*ndo_set_vf_port)(struct net_device *dev, 1227 int vf, 1228 struct nlattr *port[]); 1229 int (*ndo_get_vf_port)(struct net_device *dev, 1230 int vf, struct sk_buff *skb); 1231 int (*ndo_set_vf_guid)(struct net_device *dev, 1232 int vf, u64 guid, 1233 int guid_type); 1234 int (*ndo_set_vf_rss_query_en)( 1235 struct net_device *dev, 1236 int vf, bool setting); 1237 int (*ndo_setup_tc)(struct net_device *dev, 1238 enum tc_setup_type type, 1239 void *type_data); 1240 #if IS_ENABLED(CONFIG_FCOE) 1241 int (*ndo_fcoe_enable)(struct net_device *dev); 1242 int (*ndo_fcoe_disable)(struct net_device *dev); 1243 int (*ndo_fcoe_ddp_setup)(struct net_device *dev, 1244 u16 xid, 1245 struct scatterlist *sgl, 1246 unsigned int sgc); 1247 int (*ndo_fcoe_ddp_done)(struct net_device *dev, 1248 u16 xid); 1249 int (*ndo_fcoe_ddp_target)(struct net_device *dev, 1250 u16 xid, 1251 struct scatterlist *sgl, 1252 unsigned int sgc); 1253 int (*ndo_fcoe_get_hbainfo)(struct net_device *dev, 1254 struct netdev_fcoe_hbainfo *hbainfo); 1255 #endif 1256 1257 #if IS_ENABLED(CONFIG_LIBFCOE) 1258 #define NETDEV_FCOE_WWNN 0 1259 #define NETDEV_FCOE_WWPN 1 1260 int (*ndo_fcoe_get_wwn)(struct net_device *dev, 1261 u64 *wwn, int type); 1262 #endif 1263 1264 #ifdef CONFIG_RFS_ACCEL 1265 int (*ndo_rx_flow_steer)(struct net_device *dev, 1266 const struct sk_buff *skb, 1267 u16 rxq_index, 1268 u32 flow_id); 1269 #endif 1270 int (*ndo_add_slave)(struct net_device *dev, 1271 struct net_device *slave_dev, 1272 struct netlink_ext_ack *extack); 1273 int (*ndo_del_slave)(struct net_device *dev, 1274 struct net_device *slave_dev); 1275 netdev_features_t (*ndo_fix_features)(struct net_device *dev, 1276 netdev_features_t features); 1277 int (*ndo_set_features)(struct net_device *dev, 1278 netdev_features_t features); 1279 int (*ndo_neigh_construct)(struct net_device *dev, 1280 struct neighbour *n); 1281 void (*ndo_neigh_destroy)(struct net_device *dev, 1282 struct neighbour *n); 1283 1284 int (*ndo_fdb_add)(struct ndmsg *ndm, 1285 struct nlattr *tb[], 1286 struct net_device *dev, 1287 const unsigned char *addr, 1288 u16 vid, 1289 u16 flags); 1290 int (*ndo_fdb_del)(struct ndmsg *ndm, 1291 struct nlattr *tb[], 1292 struct net_device *dev, 1293 const unsigned char *addr, 1294 u16 vid); 1295 int (*ndo_fdb_dump)(struct sk_buff *skb, 1296 struct netlink_callback *cb, 1297 struct net_device *dev, 1298 struct net_device *filter_dev, 1299 int *idx); 1300 1301 int (*ndo_bridge_setlink)(struct net_device *dev, 1302 struct nlmsghdr *nlh, 1303 u16 flags); 1304 int (*ndo_bridge_getlink)(struct sk_buff *skb, 1305 u32 pid, u32 seq, 1306 struct net_device *dev, 1307 u32 filter_mask, 1308 int nlflags); 1309 int (*ndo_bridge_dellink)(struct net_device *dev, 1310 struct nlmsghdr *nlh, 1311 u16 flags); 1312 int (*ndo_change_carrier)(struct net_device *dev, 1313 bool new_carrier); 1314 int (*ndo_get_phys_port_id)(struct net_device *dev, 1315 struct netdev_phys_item_id *ppid); 1316 int (*ndo_get_phys_port_name)(struct net_device *dev, 1317 char *name, size_t len); 1318 void (*ndo_udp_tunnel_add)(struct net_device *dev, 1319 struct udp_tunnel_info *ti); 1320 void (*ndo_udp_tunnel_del)(struct net_device *dev, 1321 struct udp_tunnel_info *ti); 1322 void* (*ndo_dfwd_add_station)(struct net_device *pdev, 1323 struct net_device *dev); 1324 void (*ndo_dfwd_del_station)(struct net_device *pdev, 1325 void *priv); 1326 1327 int (*ndo_get_lock_subclass)(struct net_device *dev); 1328 int (*ndo_set_tx_maxrate)(struct net_device *dev, 1329 int queue_index, 1330 u32 maxrate); 1331 int (*ndo_get_iflink)(const struct net_device *dev); 1332 int (*ndo_change_proto_down)(struct net_device *dev, 1333 bool proto_down); 1334 int (*ndo_fill_metadata_dst)(struct net_device *dev, 1335 struct sk_buff *skb); 1336 void (*ndo_set_rx_headroom)(struct net_device *dev, 1337 int needed_headroom); 1338 int (*ndo_bpf)(struct net_device *dev, 1339 struct netdev_bpf *bpf); 1340 int (*ndo_xdp_xmit)(struct net_device *dev, 1341 struct xdp_buff *xdp); 1342 void (*ndo_xdp_flush)(struct net_device *dev); 1343 }; 1344 1345 /** 1346 * enum net_device_priv_flags - &struct net_device priv_flags 1347 * 1348 * These are the &struct net_device, they are only set internally 1349 * by drivers and used in the kernel. These flags are invisible to 1350 * userspace; this means that the order of these flags can change 1351 * during any kernel release. 1352 * 1353 * You should have a pretty good reason to be extending these flags. 1354 * 1355 * @IFF_802_1Q_VLAN: 802.1Q VLAN device 1356 * @IFF_EBRIDGE: Ethernet bridging device 1357 * @IFF_BONDING: bonding master or slave 1358 * @IFF_ISATAP: ISATAP interface (RFC4214) 1359 * @IFF_WAN_HDLC: WAN HDLC device 1360 * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to 1361 * release skb->dst 1362 * @IFF_DONT_BRIDGE: disallow bridging this ether dev 1363 * @IFF_DISABLE_NETPOLL: disable netpoll at run-time 1364 * @IFF_MACVLAN_PORT: device used as macvlan port 1365 * @IFF_BRIDGE_PORT: device used as bridge port 1366 * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port 1367 * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit 1368 * @IFF_UNICAST_FLT: Supports unicast filtering 1369 * @IFF_TEAM_PORT: device used as team port 1370 * @IFF_SUPP_NOFCS: device supports sending custom FCS 1371 * @IFF_LIVE_ADDR_CHANGE: device supports hardware address 1372 * change when it's running 1373 * @IFF_MACVLAN: Macvlan device 1374 * @IFF_XMIT_DST_RELEASE_PERM: IFF_XMIT_DST_RELEASE not taking into account 1375 * underlying stacked devices 1376 * @IFF_IPVLAN_MASTER: IPvlan master device 1377 * @IFF_IPVLAN_SLAVE: IPvlan slave device 1378 * @IFF_L3MDEV_MASTER: device is an L3 master device 1379 * @IFF_NO_QUEUE: device can run without qdisc attached 1380 * @IFF_OPENVSWITCH: device is a Open vSwitch master 1381 * @IFF_L3MDEV_SLAVE: device is enslaved to an L3 master device 1382 * @IFF_TEAM: device is a team device 1383 * @IFF_RXFH_CONFIGURED: device has had Rx Flow indirection table configured 1384 * @IFF_PHONY_HEADROOM: the headroom value is controlled by an external 1385 * entity (i.e. the master device for bridged veth) 1386 * @IFF_MACSEC: device is a MACsec device 1387 */ 1388 enum netdev_priv_flags { 1389 IFF_802_1Q_VLAN = 1<<0, 1390 IFF_EBRIDGE = 1<<1, 1391 IFF_BONDING = 1<<2, 1392 IFF_ISATAP = 1<<3, 1393 IFF_WAN_HDLC = 1<<4, 1394 IFF_XMIT_DST_RELEASE = 1<<5, 1395 IFF_DONT_BRIDGE = 1<<6, 1396 IFF_DISABLE_NETPOLL = 1<<7, 1397 IFF_MACVLAN_PORT = 1<<8, 1398 IFF_BRIDGE_PORT = 1<<9, 1399 IFF_OVS_DATAPATH = 1<<10, 1400 IFF_TX_SKB_SHARING = 1<<11, 1401 IFF_UNICAST_FLT = 1<<12, 1402 IFF_TEAM_PORT = 1<<13, 1403 IFF_SUPP_NOFCS = 1<<14, 1404 IFF_LIVE_ADDR_CHANGE = 1<<15, 1405 IFF_MACVLAN = 1<<16, 1406 IFF_XMIT_DST_RELEASE_PERM = 1<<17, 1407 IFF_IPVLAN_MASTER = 1<<18, 1408 IFF_IPVLAN_SLAVE = 1<<19, 1409 IFF_L3MDEV_MASTER = 1<<20, 1410 IFF_NO_QUEUE = 1<<21, 1411 IFF_OPENVSWITCH = 1<<22, 1412 IFF_L3MDEV_SLAVE = 1<<23, 1413 IFF_TEAM = 1<<24, 1414 IFF_RXFH_CONFIGURED = 1<<25, 1415 IFF_PHONY_HEADROOM = 1<<26, 1416 IFF_MACSEC = 1<<27, 1417 }; 1418 1419 #define IFF_802_1Q_VLAN IFF_802_1Q_VLAN 1420 #define IFF_EBRIDGE IFF_EBRIDGE 1421 #define IFF_BONDING IFF_BONDING 1422 #define IFF_ISATAP IFF_ISATAP 1423 #define IFF_WAN_HDLC IFF_WAN_HDLC 1424 #define IFF_XMIT_DST_RELEASE IFF_XMIT_DST_RELEASE 1425 #define IFF_DONT_BRIDGE IFF_DONT_BRIDGE 1426 #define IFF_DISABLE_NETPOLL IFF_DISABLE_NETPOLL 1427 #define IFF_MACVLAN_PORT IFF_MACVLAN_PORT 1428 #define IFF_BRIDGE_PORT IFF_BRIDGE_PORT 1429 #define IFF_OVS_DATAPATH IFF_OVS_DATAPATH 1430 #define IFF_TX_SKB_SHARING IFF_TX_SKB_SHARING 1431 #define IFF_UNICAST_FLT IFF_UNICAST_FLT 1432 #define IFF_TEAM_PORT IFF_TEAM_PORT 1433 #define IFF_SUPP_NOFCS IFF_SUPP_NOFCS 1434 #define IFF_LIVE_ADDR_CHANGE IFF_LIVE_ADDR_CHANGE 1435 #define IFF_MACVLAN IFF_MACVLAN 1436 #define IFF_XMIT_DST_RELEASE_PERM IFF_XMIT_DST_RELEASE_PERM 1437 #define IFF_IPVLAN_MASTER IFF_IPVLAN_MASTER 1438 #define IFF_IPVLAN_SLAVE IFF_IPVLAN_SLAVE 1439 #define IFF_L3MDEV_MASTER IFF_L3MDEV_MASTER 1440 #define IFF_NO_QUEUE IFF_NO_QUEUE 1441 #define IFF_OPENVSWITCH IFF_OPENVSWITCH 1442 #define IFF_L3MDEV_SLAVE IFF_L3MDEV_SLAVE 1443 #define IFF_TEAM IFF_TEAM 1444 #define IFF_RXFH_CONFIGURED IFF_RXFH_CONFIGURED 1445 #define IFF_MACSEC IFF_MACSEC 1446 1447 /** 1448 * struct net_device - The DEVICE structure. 1449 * 1450 * Actually, this whole structure is a big mistake. It mixes I/O 1451 * data with strictly "high-level" data, and it has to know about 1452 * almost every data structure used in the INET module. 1453 * 1454 * @name: This is the first field of the "visible" part of this structure 1455 * (i.e. as seen by users in the "Space.c" file). It is the name 1456 * of the interface. 1457 * 1458 * @name_hlist: Device name hash chain, please keep it close to name[] 1459 * @ifalias: SNMP alias 1460 * @mem_end: Shared memory end 1461 * @mem_start: Shared memory start 1462 * @base_addr: Device I/O address 1463 * @irq: Device IRQ number 1464 * 1465 * @carrier_changes: Stats to monitor carrier on<->off transitions 1466 * 1467 * @state: Generic network queuing layer state, see netdev_state_t 1468 * @dev_list: The global list of network devices 1469 * @napi_list: List entry used for polling NAPI devices 1470 * @unreg_list: List entry when we are unregistering the 1471 * device; see the function unregister_netdev 1472 * @close_list: List entry used when we are closing the device 1473 * @ptype_all: Device-specific packet handlers for all protocols 1474 * @ptype_specific: Device-specific, protocol-specific packet handlers 1475 * 1476 * @adj_list: Directly linked devices, like slaves for bonding 1477 * @features: Currently active device features 1478 * @hw_features: User-changeable features 1479 * 1480 * @wanted_features: User-requested features 1481 * @vlan_features: Mask of features inheritable by VLAN devices 1482 * 1483 * @hw_enc_features: Mask of features inherited by encapsulating devices 1484 * This field indicates what encapsulation 1485 * offloads the hardware is capable of doing, 1486 * and drivers will need to set them appropriately. 1487 * 1488 * @mpls_features: Mask of features inheritable by MPLS 1489 * 1490 * @ifindex: interface index 1491 * @group: The group the device belongs to 1492 * 1493 * @stats: Statistics struct, which was left as a legacy, use 1494 * rtnl_link_stats64 instead 1495 * 1496 * @rx_dropped: Dropped packets by core network, 1497 * do not use this in drivers 1498 * @tx_dropped: Dropped packets by core network, 1499 * do not use this in drivers 1500 * @rx_nohandler: nohandler dropped packets by core network on 1501 * inactive devices, do not use this in drivers 1502 * 1503 * @wireless_handlers: List of functions to handle Wireless Extensions, 1504 * instead of ioctl, 1505 * see <net/iw_handler.h> for details. 1506 * @wireless_data: Instance data managed by the core of wireless extensions 1507 * 1508 * @netdev_ops: Includes several pointers to callbacks, 1509 * if one wants to override the ndo_*() functions 1510 * @ethtool_ops: Management operations 1511 * @ndisc_ops: Includes callbacks for different IPv6 neighbour 1512 * discovery handling. Necessary for e.g. 6LoWPAN. 1513 * @header_ops: Includes callbacks for creating,parsing,caching,etc 1514 * of Layer 2 headers. 1515 * 1516 * @flags: Interface flags (a la BSD) 1517 * @priv_flags: Like 'flags' but invisible to userspace, 1518 * see if.h for the definitions 1519 * @gflags: Global flags ( kept as legacy ) 1520 * @padded: How much padding added by alloc_netdev() 1521 * @operstate: RFC2863 operstate 1522 * @link_mode: Mapping policy to operstate 1523 * @if_port: Selectable AUI, TP, ... 1524 * @dma: DMA channel 1525 * @mtu: Interface MTU value 1526 * @min_mtu: Interface Minimum MTU value 1527 * @max_mtu: Interface Maximum MTU value 1528 * @type: Interface hardware type 1529 * @hard_header_len: Maximum hardware header length. 1530 * @min_header_len: Minimum hardware header length 1531 * 1532 * @needed_headroom: Extra headroom the hardware may need, but not in all 1533 * cases can this be guaranteed 1534 * @needed_tailroom: Extra tailroom the hardware may need, but not in all 1535 * cases can this be guaranteed. Some cases also use 1536 * LL_MAX_HEADER instead to allocate the skb 1537 * 1538 * interface address info: 1539 * 1540 * @perm_addr: Permanent hw address 1541 * @addr_assign_type: Hw address assignment type 1542 * @addr_len: Hardware address length 1543 * @neigh_priv_len: Used in neigh_alloc() 1544 * @dev_id: Used to differentiate devices that share 1545 * the same link layer address 1546 * @dev_port: Used to differentiate devices that share 1547 * the same function 1548 * @addr_list_lock: XXX: need comments on this one 1549 * @uc_promisc: Counter that indicates promiscuous mode 1550 * has been enabled due to the need to listen to 1551 * additional unicast addresses in a device that 1552 * does not implement ndo_set_rx_mode() 1553 * @uc: unicast mac addresses 1554 * @mc: multicast mac addresses 1555 * @dev_addrs: list of device hw addresses 1556 * @queues_kset: Group of all Kobjects in the Tx and RX queues 1557 * @promiscuity: Number of times the NIC is told to work in 1558 * promiscuous mode; if it becomes 0 the NIC will 1559 * exit promiscuous mode 1560 * @allmulti: Counter, enables or disables allmulticast mode 1561 * 1562 * @vlan_info: VLAN info 1563 * @dsa_ptr: dsa specific data 1564 * @tipc_ptr: TIPC specific data 1565 * @atalk_ptr: AppleTalk link 1566 * @ip_ptr: IPv4 specific data 1567 * @dn_ptr: DECnet specific data 1568 * @ip6_ptr: IPv6 specific data 1569 * @ax25_ptr: AX.25 specific data 1570 * @ieee80211_ptr: IEEE 802.11 specific data, assign before registering 1571 * 1572 * @dev_addr: Hw address (before bcast, 1573 * because most packets are unicast) 1574 * 1575 * @_rx: Array of RX queues 1576 * @num_rx_queues: Number of RX queues 1577 * allocated at register_netdev() time 1578 * @real_num_rx_queues: Number of RX queues currently active in device 1579 * 1580 * @rx_handler: handler for received packets 1581 * @rx_handler_data: XXX: need comments on this one 1582 * @miniq_ingress: ingress/clsact qdisc specific data for 1583 * ingress processing 1584 * @ingress_queue: XXX: need comments on this one 1585 * @broadcast: hw bcast address 1586 * 1587 * @rx_cpu_rmap: CPU reverse-mapping for RX completion interrupts, 1588 * indexed by RX queue number. Assigned by driver. 1589 * This must only be set if the ndo_rx_flow_steer 1590 * operation is defined 1591 * @index_hlist: Device index hash chain 1592 * 1593 * @_tx: Array of TX queues 1594 * @num_tx_queues: Number of TX queues allocated at alloc_netdev_mq() time 1595 * @real_num_tx_queues: Number of TX queues currently active in device 1596 * @qdisc: Root qdisc from userspace point of view 1597 * @tx_queue_len: Max frames per queue allowed 1598 * @tx_global_lock: XXX: need comments on this one 1599 * 1600 * @xps_maps: XXX: need comments on this one 1601 * @miniq_egress: clsact qdisc specific data for 1602 * egress processing 1603 * @watchdog_timeo: Represents the timeout that is used by 1604 * the watchdog (see dev_watchdog()) 1605 * @watchdog_timer: List of timers 1606 * 1607 * @pcpu_refcnt: Number of references to this device 1608 * @todo_list: Delayed register/unregister 1609 * @link_watch_list: XXX: need comments on this one 1610 * 1611 * @reg_state: Register/unregister state machine 1612 * @dismantle: Device is going to be freed 1613 * @rtnl_link_state: This enum represents the phases of creating 1614 * a new link 1615 * 1616 * @needs_free_netdev: Should unregister perform free_netdev? 1617 * @priv_destructor: Called from unregister 1618 * @npinfo: XXX: need comments on this one 1619 * @nd_net: Network namespace this network device is inside 1620 * 1621 * @ml_priv: Mid-layer private 1622 * @lstats: Loopback statistics 1623 * @tstats: Tunnel statistics 1624 * @dstats: Dummy statistics 1625 * @vstats: Virtual ethernet statistics 1626 * 1627 * @garp_port: GARP 1628 * @mrp_port: MRP 1629 * 1630 * @dev: Class/net/name entry 1631 * @sysfs_groups: Space for optional device, statistics and wireless 1632 * sysfs groups 1633 * 1634 * @sysfs_rx_queue_group: Space for optional per-rx queue attributes 1635 * @rtnl_link_ops: Rtnl_link_ops 1636 * 1637 * @gso_max_size: Maximum size of generic segmentation offload 1638 * @gso_max_segs: Maximum number of segments that can be passed to the 1639 * NIC for GSO 1640 * 1641 * @dcbnl_ops: Data Center Bridging netlink ops 1642 * @num_tc: Number of traffic classes in the net device 1643 * @tc_to_txq: XXX: need comments on this one 1644 * @prio_tc_map: XXX: need comments on this one 1645 * 1646 * @fcoe_ddp_xid: Max exchange id for FCoE LRO by ddp 1647 * 1648 * @priomap: XXX: need comments on this one 1649 * @phydev: Physical device may attach itself 1650 * for hardware timestamping 1651 * 1652 * @qdisc_tx_busylock: lockdep class annotating Qdisc->busylock spinlock 1653 * @qdisc_running_key: lockdep class annotating Qdisc->running seqcount 1654 * 1655 * @proto_down: protocol port state information can be sent to the 1656 * switch driver and used to set the phys state of the 1657 * switch port. 1658 * 1659 * FIXME: cleanup struct net_device such that network protocol info 1660 * moves out. 1661 */ 1662 1663 struct net_device { 1664 char name[IFNAMSIZ]; 1665 struct hlist_node name_hlist; 1666 struct dev_ifalias __rcu *ifalias; 1667 /* 1668 * I/O specific fields 1669 * FIXME: Merge these and struct ifmap into one 1670 */ 1671 unsigned long mem_end; 1672 unsigned long mem_start; 1673 unsigned long base_addr; 1674 int irq; 1675 1676 atomic_t carrier_changes; 1677 1678 /* 1679 * Some hardware also needs these fields (state,dev_list, 1680 * napi_list,unreg_list,close_list) but they are not 1681 * part of the usual set specified in Space.c. 1682 */ 1683 1684 unsigned long state; 1685 1686 struct list_head dev_list; 1687 struct list_head napi_list; 1688 struct list_head unreg_list; 1689 struct list_head close_list; 1690 struct list_head ptype_all; 1691 struct list_head ptype_specific; 1692 1693 struct { 1694 struct list_head upper; 1695 struct list_head lower; 1696 } adj_list; 1697 1698 netdev_features_t features; 1699 netdev_features_t hw_features; 1700 netdev_features_t wanted_features; 1701 netdev_features_t vlan_features; 1702 netdev_features_t hw_enc_features; 1703 netdev_features_t mpls_features; 1704 netdev_features_t gso_partial_features; 1705 1706 int ifindex; 1707 int group; 1708 1709 struct net_device_stats stats; 1710 1711 atomic_long_t rx_dropped; 1712 atomic_long_t tx_dropped; 1713 atomic_long_t rx_nohandler; 1714 1715 #ifdef CONFIG_WIRELESS_EXT 1716 const struct iw_handler_def *wireless_handlers; 1717 struct iw_public_data *wireless_data; 1718 #endif 1719 const struct net_device_ops *netdev_ops; 1720 const struct ethtool_ops *ethtool_ops; 1721 #ifdef CONFIG_NET_SWITCHDEV 1722 const struct switchdev_ops *switchdev_ops; 1723 #endif 1724 #ifdef CONFIG_NET_L3_MASTER_DEV 1725 const struct l3mdev_ops *l3mdev_ops; 1726 #endif 1727 #if IS_ENABLED(CONFIG_IPV6) 1728 const struct ndisc_ops *ndisc_ops; 1729 #endif 1730 1731 #ifdef CONFIG_XFRM_OFFLOAD 1732 const struct xfrmdev_ops *xfrmdev_ops; 1733 #endif 1734 1735 const struct header_ops *header_ops; 1736 1737 unsigned int flags; 1738 unsigned int priv_flags; 1739 1740 unsigned short gflags; 1741 unsigned short padded; 1742 1743 unsigned char operstate; 1744 unsigned char link_mode; 1745 1746 unsigned char if_port; 1747 unsigned char dma; 1748 1749 unsigned int mtu; 1750 unsigned int min_mtu; 1751 unsigned int max_mtu; 1752 unsigned short type; 1753 unsigned short hard_header_len; 1754 unsigned char min_header_len; 1755 1756 unsigned short needed_headroom; 1757 unsigned short needed_tailroom; 1758 1759 /* Interface address info. */ 1760 unsigned char perm_addr[MAX_ADDR_LEN]; 1761 unsigned char addr_assign_type; 1762 unsigned char addr_len; 1763 unsigned short neigh_priv_len; 1764 unsigned short dev_id; 1765 unsigned short dev_port; 1766 spinlock_t addr_list_lock; 1767 unsigned char name_assign_type; 1768 bool uc_promisc; 1769 struct netdev_hw_addr_list uc; 1770 struct netdev_hw_addr_list mc; 1771 struct netdev_hw_addr_list dev_addrs; 1772 1773 #ifdef CONFIG_SYSFS 1774 struct kset *queues_kset; 1775 #endif 1776 unsigned int promiscuity; 1777 unsigned int allmulti; 1778 1779 1780 /* Protocol-specific pointers */ 1781 1782 #if IS_ENABLED(CONFIG_VLAN_8021Q) 1783 struct vlan_info __rcu *vlan_info; 1784 #endif 1785 #if IS_ENABLED(CONFIG_NET_DSA) 1786 struct dsa_port *dsa_ptr; 1787 #endif 1788 #if IS_ENABLED(CONFIG_TIPC) 1789 struct tipc_bearer __rcu *tipc_ptr; 1790 #endif 1791 void *atalk_ptr; 1792 struct in_device __rcu *ip_ptr; 1793 struct dn_dev __rcu *dn_ptr; 1794 struct inet6_dev __rcu *ip6_ptr; 1795 void *ax25_ptr; 1796 struct wireless_dev *ieee80211_ptr; 1797 struct wpan_dev *ieee802154_ptr; 1798 #if IS_ENABLED(CONFIG_MPLS_ROUTING) 1799 struct mpls_dev __rcu *mpls_ptr; 1800 #endif 1801 1802 /* 1803 * Cache lines mostly used on receive path (including eth_type_trans()) 1804 */ 1805 /* Interface address info used in eth_type_trans() */ 1806 unsigned char *dev_addr; 1807 1808 #ifdef CONFIG_SYSFS 1809 struct netdev_rx_queue *_rx; 1810 1811 unsigned int num_rx_queues; 1812 unsigned int real_num_rx_queues; 1813 #endif 1814 1815 struct bpf_prog __rcu *xdp_prog; 1816 unsigned long gro_flush_timeout; 1817 rx_handler_func_t __rcu *rx_handler; 1818 void __rcu *rx_handler_data; 1819 1820 #ifdef CONFIG_NET_CLS_ACT 1821 struct mini_Qdisc __rcu *miniq_ingress; 1822 #endif 1823 struct netdev_queue __rcu *ingress_queue; 1824 #ifdef CONFIG_NETFILTER_INGRESS 1825 struct nf_hook_entries __rcu *nf_hooks_ingress; 1826 #endif 1827 1828 unsigned char broadcast[MAX_ADDR_LEN]; 1829 #ifdef CONFIG_RFS_ACCEL 1830 struct cpu_rmap *rx_cpu_rmap; 1831 #endif 1832 struct hlist_node index_hlist; 1833 1834 /* 1835 * Cache lines mostly used on transmit path 1836 */ 1837 struct netdev_queue *_tx ____cacheline_aligned_in_smp; 1838 unsigned int num_tx_queues; 1839 unsigned int real_num_tx_queues; 1840 struct Qdisc *qdisc; 1841 #ifdef CONFIG_NET_SCHED 1842 DECLARE_HASHTABLE (qdisc_hash, 4); 1843 #endif 1844 unsigned int tx_queue_len; 1845 spinlock_t tx_global_lock; 1846 int watchdog_timeo; 1847 1848 #ifdef CONFIG_XPS 1849 struct xps_dev_maps __rcu *xps_maps; 1850 #endif 1851 #ifdef CONFIG_NET_CLS_ACT 1852 struct mini_Qdisc __rcu *miniq_egress; 1853 #endif 1854 1855 /* These may be needed for future network-power-down code. */ 1856 struct timer_list watchdog_timer; 1857 1858 int __percpu *pcpu_refcnt; 1859 struct list_head todo_list; 1860 1861 struct list_head link_watch_list; 1862 1863 enum { NETREG_UNINITIALIZED=0, 1864 NETREG_REGISTERED, /* completed register_netdevice */ 1865 NETREG_UNREGISTERING, /* called unregister_netdevice */ 1866 NETREG_UNREGISTERED, /* completed unregister todo */ 1867 NETREG_RELEASED, /* called free_netdev */ 1868 NETREG_DUMMY, /* dummy device for NAPI poll */ 1869 } reg_state:8; 1870 1871 bool dismantle; 1872 1873 enum { 1874 RTNL_LINK_INITIALIZED, 1875 RTNL_LINK_INITIALIZING, 1876 } rtnl_link_state:16; 1877 1878 bool needs_free_netdev; 1879 void (*priv_destructor)(struct net_device *dev); 1880 1881 #ifdef CONFIG_NETPOLL 1882 struct netpoll_info __rcu *npinfo; 1883 #endif 1884 1885 possible_net_t nd_net; 1886 1887 /* mid-layer private */ 1888 union { 1889 void *ml_priv; 1890 struct pcpu_lstats __percpu *lstats; 1891 struct pcpu_sw_netstats __percpu *tstats; 1892 struct pcpu_dstats __percpu *dstats; 1893 struct pcpu_vstats __percpu *vstats; 1894 }; 1895 1896 #if IS_ENABLED(CONFIG_GARP) 1897 struct garp_port __rcu *garp_port; 1898 #endif 1899 #if IS_ENABLED(CONFIG_MRP) 1900 struct mrp_port __rcu *mrp_port; 1901 #endif 1902 1903 struct device dev; 1904 const struct attribute_group *sysfs_groups[4]; 1905 const struct attribute_group *sysfs_rx_queue_group; 1906 1907 const struct rtnl_link_ops *rtnl_link_ops; 1908 1909 /* for setting kernel sock attribute on TCP connection setup */ 1910 #define GSO_MAX_SIZE 65536 1911 unsigned int gso_max_size; 1912 #define GSO_MAX_SEGS 65535 1913 u16 gso_max_segs; 1914 1915 #ifdef CONFIG_DCB 1916 const struct dcbnl_rtnl_ops *dcbnl_ops; 1917 #endif 1918 u8 num_tc; 1919 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE]; 1920 u8 prio_tc_map[TC_BITMASK + 1]; 1921 1922 #if IS_ENABLED(CONFIG_FCOE) 1923 unsigned int fcoe_ddp_xid; 1924 #endif 1925 #if IS_ENABLED(CONFIG_CGROUP_NET_PRIO) 1926 struct netprio_map __rcu *priomap; 1927 #endif 1928 struct phy_device *phydev; 1929 struct lock_class_key *qdisc_tx_busylock; 1930 struct lock_class_key *qdisc_running_key; 1931 bool proto_down; 1932 }; 1933 #define to_net_dev(d) container_of(d, struct net_device, dev) 1934 1935 static inline bool netif_elide_gro(const struct net_device *dev) 1936 { 1937 if (!(dev->features & NETIF_F_GRO) || dev->xdp_prog) 1938 return true; 1939 return false; 1940 } 1941 1942 #define NETDEV_ALIGN 32 1943 1944 static inline 1945 int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio) 1946 { 1947 return dev->prio_tc_map[prio & TC_BITMASK]; 1948 } 1949 1950 static inline 1951 int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc) 1952 { 1953 if (tc >= dev->num_tc) 1954 return -EINVAL; 1955 1956 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK; 1957 return 0; 1958 } 1959 1960 int netdev_txq_to_tc(struct net_device *dev, unsigned int txq); 1961 void netdev_reset_tc(struct net_device *dev); 1962 int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset); 1963 int netdev_set_num_tc(struct net_device *dev, u8 num_tc); 1964 1965 static inline 1966 int netdev_get_num_tc(struct net_device *dev) 1967 { 1968 return dev->num_tc; 1969 } 1970 1971 static inline 1972 struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev, 1973 unsigned int index) 1974 { 1975 return &dev->_tx[index]; 1976 } 1977 1978 static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev, 1979 const struct sk_buff *skb) 1980 { 1981 return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb)); 1982 } 1983 1984 static inline void netdev_for_each_tx_queue(struct net_device *dev, 1985 void (*f)(struct net_device *, 1986 struct netdev_queue *, 1987 void *), 1988 void *arg) 1989 { 1990 unsigned int i; 1991 1992 for (i = 0; i < dev->num_tx_queues; i++) 1993 f(dev, &dev->_tx[i], arg); 1994 } 1995 1996 #define netdev_lockdep_set_classes(dev) \ 1997 { \ 1998 static struct lock_class_key qdisc_tx_busylock_key; \ 1999 static struct lock_class_key qdisc_running_key; \ 2000 static struct lock_class_key qdisc_xmit_lock_key; \ 2001 static struct lock_class_key dev_addr_list_lock_key; \ 2002 unsigned int i; \ 2003 \ 2004 (dev)->qdisc_tx_busylock = &qdisc_tx_busylock_key; \ 2005 (dev)->qdisc_running_key = &qdisc_running_key; \ 2006 lockdep_set_class(&(dev)->addr_list_lock, \ 2007 &dev_addr_list_lock_key); \ 2008 for (i = 0; i < (dev)->num_tx_queues; i++) \ 2009 lockdep_set_class(&(dev)->_tx[i]._xmit_lock, \ 2010 &qdisc_xmit_lock_key); \ 2011 } 2012 2013 struct netdev_queue *netdev_pick_tx(struct net_device *dev, 2014 struct sk_buff *skb, 2015 void *accel_priv); 2016 2017 /* returns the headroom that the master device needs to take in account 2018 * when forwarding to this dev 2019 */ 2020 static inline unsigned netdev_get_fwd_headroom(struct net_device *dev) 2021 { 2022 return dev->priv_flags & IFF_PHONY_HEADROOM ? 0 : dev->needed_headroom; 2023 } 2024 2025 static inline void netdev_set_rx_headroom(struct net_device *dev, int new_hr) 2026 { 2027 if (dev->netdev_ops->ndo_set_rx_headroom) 2028 dev->netdev_ops->ndo_set_rx_headroom(dev, new_hr); 2029 } 2030 2031 /* set the device rx headroom to the dev's default */ 2032 static inline void netdev_reset_rx_headroom(struct net_device *dev) 2033 { 2034 netdev_set_rx_headroom(dev, -1); 2035 } 2036 2037 /* 2038 * Net namespace inlines 2039 */ 2040 static inline 2041 struct net *dev_net(const struct net_device *dev) 2042 { 2043 return read_pnet(&dev->nd_net); 2044 } 2045 2046 static inline 2047 void dev_net_set(struct net_device *dev, struct net *net) 2048 { 2049 write_pnet(&dev->nd_net, net); 2050 } 2051 2052 /** 2053 * netdev_priv - access network device private data 2054 * @dev: network device 2055 * 2056 * Get network device private data 2057 */ 2058 static inline void *netdev_priv(const struct net_device *dev) 2059 { 2060 return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN); 2061 } 2062 2063 /* Set the sysfs physical device reference for the network logical device 2064 * if set prior to registration will cause a symlink during initialization. 2065 */ 2066 #define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev)) 2067 2068 /* Set the sysfs device type for the network logical device to allow 2069 * fine-grained identification of different network device types. For 2070 * example Ethernet, Wireless LAN, Bluetooth, WiMAX etc. 2071 */ 2072 #define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype)) 2073 2074 /* Default NAPI poll() weight 2075 * Device drivers are strongly advised to not use bigger value 2076 */ 2077 #define NAPI_POLL_WEIGHT 64 2078 2079 /** 2080 * netif_napi_add - initialize a NAPI context 2081 * @dev: network device 2082 * @napi: NAPI context 2083 * @poll: polling function 2084 * @weight: default weight 2085 * 2086 * netif_napi_add() must be used to initialize a NAPI context prior to calling 2087 * *any* of the other NAPI-related functions. 2088 */ 2089 void netif_napi_add(struct net_device *dev, struct napi_struct *napi, 2090 int (*poll)(struct napi_struct *, int), int weight); 2091 2092 /** 2093 * netif_tx_napi_add - initialize a NAPI context 2094 * @dev: network device 2095 * @napi: NAPI context 2096 * @poll: polling function 2097 * @weight: default weight 2098 * 2099 * This variant of netif_napi_add() should be used from drivers using NAPI 2100 * to exclusively poll a TX queue. 2101 * This will avoid we add it into napi_hash[], thus polluting this hash table. 2102 */ 2103 static inline void netif_tx_napi_add(struct net_device *dev, 2104 struct napi_struct *napi, 2105 int (*poll)(struct napi_struct *, int), 2106 int weight) 2107 { 2108 set_bit(NAPI_STATE_NO_BUSY_POLL, &napi->state); 2109 netif_napi_add(dev, napi, poll, weight); 2110 } 2111 2112 /** 2113 * netif_napi_del - remove a NAPI context 2114 * @napi: NAPI context 2115 * 2116 * netif_napi_del() removes a NAPI context from the network device NAPI list 2117 */ 2118 void netif_napi_del(struct napi_struct *napi); 2119 2120 struct napi_gro_cb { 2121 /* Virtual address of skb_shinfo(skb)->frags[0].page + offset. */ 2122 void *frag0; 2123 2124 /* Length of frag0. */ 2125 unsigned int frag0_len; 2126 2127 /* This indicates where we are processing relative to skb->data. */ 2128 int data_offset; 2129 2130 /* This is non-zero if the packet cannot be merged with the new skb. */ 2131 u16 flush; 2132 2133 /* Save the IP ID here and check when we get to the transport layer */ 2134 u16 flush_id; 2135 2136 /* Number of segments aggregated. */ 2137 u16 count; 2138 2139 /* Start offset for remote checksum offload */ 2140 u16 gro_remcsum_start; 2141 2142 /* jiffies when first packet was created/queued */ 2143 unsigned long age; 2144 2145 /* Used in ipv6_gro_receive() and foo-over-udp */ 2146 u16 proto; 2147 2148 /* This is non-zero if the packet may be of the same flow. */ 2149 u8 same_flow:1; 2150 2151 /* Used in tunnel GRO receive */ 2152 u8 encap_mark:1; 2153 2154 /* GRO checksum is valid */ 2155 u8 csum_valid:1; 2156 2157 /* Number of checksums via CHECKSUM_UNNECESSARY */ 2158 u8 csum_cnt:3; 2159 2160 /* Free the skb? */ 2161 u8 free:2; 2162 #define NAPI_GRO_FREE 1 2163 #define NAPI_GRO_FREE_STOLEN_HEAD 2 2164 2165 /* Used in foo-over-udp, set in udp[46]_gro_receive */ 2166 u8 is_ipv6:1; 2167 2168 /* Used in GRE, set in fou/gue_gro_receive */ 2169 u8 is_fou:1; 2170 2171 /* Used to determine if flush_id can be ignored */ 2172 u8 is_atomic:1; 2173 2174 /* Number of gro_receive callbacks this packet already went through */ 2175 u8 recursion_counter:4; 2176 2177 /* 1 bit hole */ 2178 2179 /* used to support CHECKSUM_COMPLETE for tunneling protocols */ 2180 __wsum csum; 2181 2182 /* used in skb_gro_receive() slow path */ 2183 struct sk_buff *last; 2184 }; 2185 2186 #define NAPI_GRO_CB(skb) ((struct napi_gro_cb *)(skb)->cb) 2187 2188 #define GRO_RECURSION_LIMIT 15 2189 static inline int gro_recursion_inc_test(struct sk_buff *skb) 2190 { 2191 return ++NAPI_GRO_CB(skb)->recursion_counter == GRO_RECURSION_LIMIT; 2192 } 2193 2194 typedef struct sk_buff **(*gro_receive_t)(struct sk_buff **, struct sk_buff *); 2195 static inline struct sk_buff **call_gro_receive(gro_receive_t cb, 2196 struct sk_buff **head, 2197 struct sk_buff *skb) 2198 { 2199 if (unlikely(gro_recursion_inc_test(skb))) { 2200 NAPI_GRO_CB(skb)->flush |= 1; 2201 return NULL; 2202 } 2203 2204 return cb(head, skb); 2205 } 2206 2207 typedef struct sk_buff **(*gro_receive_sk_t)(struct sock *, struct sk_buff **, 2208 struct sk_buff *); 2209 static inline struct sk_buff **call_gro_receive_sk(gro_receive_sk_t cb, 2210 struct sock *sk, 2211 struct sk_buff **head, 2212 struct sk_buff *skb) 2213 { 2214 if (unlikely(gro_recursion_inc_test(skb))) { 2215 NAPI_GRO_CB(skb)->flush |= 1; 2216 return NULL; 2217 } 2218 2219 return cb(sk, head, skb); 2220 } 2221 2222 struct packet_type { 2223 __be16 type; /* This is really htons(ether_type). */ 2224 struct net_device *dev; /* NULL is wildcarded here */ 2225 int (*func) (struct sk_buff *, 2226 struct net_device *, 2227 struct packet_type *, 2228 struct net_device *); 2229 bool (*id_match)(struct packet_type *ptype, 2230 struct sock *sk); 2231 void *af_packet_priv; 2232 struct list_head list; 2233 }; 2234 2235 struct offload_callbacks { 2236 struct sk_buff *(*gso_segment)(struct sk_buff *skb, 2237 netdev_features_t features); 2238 struct sk_buff **(*gro_receive)(struct sk_buff **head, 2239 struct sk_buff *skb); 2240 int (*gro_complete)(struct sk_buff *skb, int nhoff); 2241 }; 2242 2243 struct packet_offload { 2244 __be16 type; /* This is really htons(ether_type). */ 2245 u16 priority; 2246 struct offload_callbacks callbacks; 2247 struct list_head list; 2248 }; 2249 2250 /* often modified stats are per-CPU, other are shared (netdev->stats) */ 2251 struct pcpu_sw_netstats { 2252 u64 rx_packets; 2253 u64 rx_bytes; 2254 u64 tx_packets; 2255 u64 tx_bytes; 2256 struct u64_stats_sync syncp; 2257 }; 2258 2259 #define __netdev_alloc_pcpu_stats(type, gfp) \ 2260 ({ \ 2261 typeof(type) __percpu *pcpu_stats = alloc_percpu_gfp(type, gfp);\ 2262 if (pcpu_stats) { \ 2263 int __cpu; \ 2264 for_each_possible_cpu(__cpu) { \ 2265 typeof(type) *stat; \ 2266 stat = per_cpu_ptr(pcpu_stats, __cpu); \ 2267 u64_stats_init(&stat->syncp); \ 2268 } \ 2269 } \ 2270 pcpu_stats; \ 2271 }) 2272 2273 #define netdev_alloc_pcpu_stats(type) \ 2274 __netdev_alloc_pcpu_stats(type, GFP_KERNEL) 2275 2276 enum netdev_lag_tx_type { 2277 NETDEV_LAG_TX_TYPE_UNKNOWN, 2278 NETDEV_LAG_TX_TYPE_RANDOM, 2279 NETDEV_LAG_TX_TYPE_BROADCAST, 2280 NETDEV_LAG_TX_TYPE_ROUNDROBIN, 2281 NETDEV_LAG_TX_TYPE_ACTIVEBACKUP, 2282 NETDEV_LAG_TX_TYPE_HASH, 2283 }; 2284 2285 struct netdev_lag_upper_info { 2286 enum netdev_lag_tx_type tx_type; 2287 }; 2288 2289 struct netdev_lag_lower_state_info { 2290 u8 link_up : 1, 2291 tx_enabled : 1; 2292 }; 2293 2294 #include <linux/notifier.h> 2295 2296 /* netdevice notifier chain. Please remember to update the rtnetlink 2297 * notification exclusion list in rtnetlink_event() when adding new 2298 * types. 2299 */ 2300 #define NETDEV_UP 0x0001 /* For now you can't veto a device up/down */ 2301 #define NETDEV_DOWN 0x0002 2302 #define NETDEV_REBOOT 0x0003 /* Tell a protocol stack a network interface 2303 detected a hardware crash and restarted 2304 - we can use this eg to kick tcp sessions 2305 once done */ 2306 #define NETDEV_CHANGE 0x0004 /* Notify device state change */ 2307 #define NETDEV_REGISTER 0x0005 2308 #define NETDEV_UNREGISTER 0x0006 2309 #define NETDEV_CHANGEMTU 0x0007 /* notify after mtu change happened */ 2310 #define NETDEV_CHANGEADDR 0x0008 2311 #define NETDEV_GOING_DOWN 0x0009 2312 #define NETDEV_CHANGENAME 0x000A 2313 #define NETDEV_FEAT_CHANGE 0x000B 2314 #define NETDEV_BONDING_FAILOVER 0x000C 2315 #define NETDEV_PRE_UP 0x000D 2316 #define NETDEV_PRE_TYPE_CHANGE 0x000E 2317 #define NETDEV_POST_TYPE_CHANGE 0x000F 2318 #define NETDEV_POST_INIT 0x0010 2319 #define NETDEV_UNREGISTER_FINAL 0x0011 2320 #define NETDEV_RELEASE 0x0012 2321 #define NETDEV_NOTIFY_PEERS 0x0013 2322 #define NETDEV_JOIN 0x0014 2323 #define NETDEV_CHANGEUPPER 0x0015 2324 #define NETDEV_RESEND_IGMP 0x0016 2325 #define NETDEV_PRECHANGEMTU 0x0017 /* notify before mtu change happened */ 2326 #define NETDEV_CHANGEINFODATA 0x0018 2327 #define NETDEV_BONDING_INFO 0x0019 2328 #define NETDEV_PRECHANGEUPPER 0x001A 2329 #define NETDEV_CHANGELOWERSTATE 0x001B 2330 #define NETDEV_UDP_TUNNEL_PUSH_INFO 0x001C 2331 #define NETDEV_UDP_TUNNEL_DROP_INFO 0x001D 2332 #define NETDEV_CHANGE_TX_QUEUE_LEN 0x001E 2333 2334 int register_netdevice_notifier(struct notifier_block *nb); 2335 int unregister_netdevice_notifier(struct notifier_block *nb); 2336 2337 struct netdev_notifier_info { 2338 struct net_device *dev; 2339 struct netlink_ext_ack *extack; 2340 }; 2341 2342 struct netdev_notifier_change_info { 2343 struct netdev_notifier_info info; /* must be first */ 2344 unsigned int flags_changed; 2345 }; 2346 2347 struct netdev_notifier_changeupper_info { 2348 struct netdev_notifier_info info; /* must be first */ 2349 struct net_device *upper_dev; /* new upper dev */ 2350 bool master; /* is upper dev master */ 2351 bool linking; /* is the notification for link or unlink */ 2352 void *upper_info; /* upper dev info */ 2353 }; 2354 2355 struct netdev_notifier_changelowerstate_info { 2356 struct netdev_notifier_info info; /* must be first */ 2357 void *lower_state_info; /* is lower dev state */ 2358 }; 2359 2360 static inline void netdev_notifier_info_init(struct netdev_notifier_info *info, 2361 struct net_device *dev) 2362 { 2363 info->dev = dev; 2364 info->extack = NULL; 2365 } 2366 2367 static inline struct net_device * 2368 netdev_notifier_info_to_dev(const struct netdev_notifier_info *info) 2369 { 2370 return info->dev; 2371 } 2372 2373 static inline struct netlink_ext_ack * 2374 netdev_notifier_info_to_extack(const struct netdev_notifier_info *info) 2375 { 2376 return info->extack; 2377 } 2378 2379 int call_netdevice_notifiers(unsigned long val, struct net_device *dev); 2380 2381 2382 extern rwlock_t dev_base_lock; /* Device list lock */ 2383 2384 #define for_each_netdev(net, d) \ 2385 list_for_each_entry(d, &(net)->dev_base_head, dev_list) 2386 #define for_each_netdev_reverse(net, d) \ 2387 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list) 2388 #define for_each_netdev_rcu(net, d) \ 2389 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list) 2390 #define for_each_netdev_safe(net, d, n) \ 2391 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list) 2392 #define for_each_netdev_continue(net, d) \ 2393 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list) 2394 #define for_each_netdev_continue_rcu(net, d) \ 2395 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list) 2396 #define for_each_netdev_in_bond_rcu(bond, slave) \ 2397 for_each_netdev_rcu(&init_net, slave) \ 2398 if (netdev_master_upper_dev_get_rcu(slave) == (bond)) 2399 #define net_device_entry(lh) list_entry(lh, struct net_device, dev_list) 2400 2401 static inline struct net_device *next_net_device(struct net_device *dev) 2402 { 2403 struct list_head *lh; 2404 struct net *net; 2405 2406 net = dev_net(dev); 2407 lh = dev->dev_list.next; 2408 return lh == &net->dev_base_head ? NULL : net_device_entry(lh); 2409 } 2410 2411 static inline struct net_device *next_net_device_rcu(struct net_device *dev) 2412 { 2413 struct list_head *lh; 2414 struct net *net; 2415 2416 net = dev_net(dev); 2417 lh = rcu_dereference(list_next_rcu(&dev->dev_list)); 2418 return lh == &net->dev_base_head ? NULL : net_device_entry(lh); 2419 } 2420 2421 static inline struct net_device *first_net_device(struct net *net) 2422 { 2423 return list_empty(&net->dev_base_head) ? NULL : 2424 net_device_entry(net->dev_base_head.next); 2425 } 2426 2427 static inline struct net_device *first_net_device_rcu(struct net *net) 2428 { 2429 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head)); 2430 2431 return lh == &net->dev_base_head ? NULL : net_device_entry(lh); 2432 } 2433 2434 int netdev_boot_setup_check(struct net_device *dev); 2435 unsigned long netdev_boot_base(const char *prefix, int unit); 2436 struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type, 2437 const char *hwaddr); 2438 struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type); 2439 struct net_device *__dev_getfirstbyhwtype(struct net *net, unsigned short type); 2440 void dev_add_pack(struct packet_type *pt); 2441 void dev_remove_pack(struct packet_type *pt); 2442 void __dev_remove_pack(struct packet_type *pt); 2443 void dev_add_offload(struct packet_offload *po); 2444 void dev_remove_offload(struct packet_offload *po); 2445 2446 int dev_get_iflink(const struct net_device *dev); 2447 int dev_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb); 2448 struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags, 2449 unsigned short mask); 2450 struct net_device *dev_get_by_name(struct net *net, const char *name); 2451 struct net_device *dev_get_by_name_rcu(struct net *net, const char *name); 2452 struct net_device *__dev_get_by_name(struct net *net, const char *name); 2453 int dev_alloc_name(struct net_device *dev, const char *name); 2454 int dev_open(struct net_device *dev); 2455 void dev_close(struct net_device *dev); 2456 void dev_close_many(struct list_head *head, bool unlink); 2457 void dev_disable_lro(struct net_device *dev); 2458 int dev_loopback_xmit(struct net *net, struct sock *sk, struct sk_buff *newskb); 2459 int dev_queue_xmit(struct sk_buff *skb); 2460 int dev_queue_xmit_accel(struct sk_buff *skb, void *accel_priv); 2461 int register_netdevice(struct net_device *dev); 2462 void unregister_netdevice_queue(struct net_device *dev, struct list_head *head); 2463 void unregister_netdevice_many(struct list_head *head); 2464 static inline void unregister_netdevice(struct net_device *dev) 2465 { 2466 unregister_netdevice_queue(dev, NULL); 2467 } 2468 2469 int netdev_refcnt_read(const struct net_device *dev); 2470 void free_netdev(struct net_device *dev); 2471 void netdev_freemem(struct net_device *dev); 2472 void synchronize_net(void); 2473 int init_dummy_netdev(struct net_device *dev); 2474 2475 DECLARE_PER_CPU(int, xmit_recursion); 2476 #define XMIT_RECURSION_LIMIT 10 2477 2478 static inline int dev_recursion_level(void) 2479 { 2480 return this_cpu_read(xmit_recursion); 2481 } 2482 2483 struct net_device *dev_get_by_index(struct net *net, int ifindex); 2484 struct net_device *__dev_get_by_index(struct net *net, int ifindex); 2485 struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex); 2486 struct net_device *dev_get_by_napi_id(unsigned int napi_id); 2487 int netdev_get_name(struct net *net, char *name, int ifindex); 2488 int dev_restart(struct net_device *dev); 2489 int skb_gro_receive(struct sk_buff **head, struct sk_buff *skb); 2490 2491 static inline unsigned int skb_gro_offset(const struct sk_buff *skb) 2492 { 2493 return NAPI_GRO_CB(skb)->data_offset; 2494 } 2495 2496 static inline unsigned int skb_gro_len(const struct sk_buff *skb) 2497 { 2498 return skb->len - NAPI_GRO_CB(skb)->data_offset; 2499 } 2500 2501 static inline void skb_gro_pull(struct sk_buff *skb, unsigned int len) 2502 { 2503 NAPI_GRO_CB(skb)->data_offset += len; 2504 } 2505 2506 static inline void *skb_gro_header_fast(struct sk_buff *skb, 2507 unsigned int offset) 2508 { 2509 return NAPI_GRO_CB(skb)->frag0 + offset; 2510 } 2511 2512 static inline int skb_gro_header_hard(struct sk_buff *skb, unsigned int hlen) 2513 { 2514 return NAPI_GRO_CB(skb)->frag0_len < hlen; 2515 } 2516 2517 static inline void skb_gro_frag0_invalidate(struct sk_buff *skb) 2518 { 2519 NAPI_GRO_CB(skb)->frag0 = NULL; 2520 NAPI_GRO_CB(skb)->frag0_len = 0; 2521 } 2522 2523 static inline void *skb_gro_header_slow(struct sk_buff *skb, unsigned int hlen, 2524 unsigned int offset) 2525 { 2526 if (!pskb_may_pull(skb, hlen)) 2527 return NULL; 2528 2529 skb_gro_frag0_invalidate(skb); 2530 return skb->data + offset; 2531 } 2532 2533 static inline void *skb_gro_network_header(struct sk_buff *skb) 2534 { 2535 return (NAPI_GRO_CB(skb)->frag0 ?: skb->data) + 2536 skb_network_offset(skb); 2537 } 2538 2539 static inline void skb_gro_postpull_rcsum(struct sk_buff *skb, 2540 const void *start, unsigned int len) 2541 { 2542 if (NAPI_GRO_CB(skb)->csum_valid) 2543 NAPI_GRO_CB(skb)->csum = csum_sub(NAPI_GRO_CB(skb)->csum, 2544 csum_partial(start, len, 0)); 2545 } 2546 2547 /* GRO checksum functions. These are logical equivalents of the normal 2548 * checksum functions (in skbuff.h) except that they operate on the GRO 2549 * offsets and fields in sk_buff. 2550 */ 2551 2552 __sum16 __skb_gro_checksum_complete(struct sk_buff *skb); 2553 2554 static inline bool skb_at_gro_remcsum_start(struct sk_buff *skb) 2555 { 2556 return (NAPI_GRO_CB(skb)->gro_remcsum_start == skb_gro_offset(skb)); 2557 } 2558 2559 static inline bool __skb_gro_checksum_validate_needed(struct sk_buff *skb, 2560 bool zero_okay, 2561 __sum16 check) 2562 { 2563 return ((skb->ip_summed != CHECKSUM_PARTIAL || 2564 skb_checksum_start_offset(skb) < 2565 skb_gro_offset(skb)) && 2566 !skb_at_gro_remcsum_start(skb) && 2567 NAPI_GRO_CB(skb)->csum_cnt == 0 && 2568 (!zero_okay || check)); 2569 } 2570 2571 static inline __sum16 __skb_gro_checksum_validate_complete(struct sk_buff *skb, 2572 __wsum psum) 2573 { 2574 if (NAPI_GRO_CB(skb)->csum_valid && 2575 !csum_fold(csum_add(psum, NAPI_GRO_CB(skb)->csum))) 2576 return 0; 2577 2578 NAPI_GRO_CB(skb)->csum = psum; 2579 2580 return __skb_gro_checksum_complete(skb); 2581 } 2582 2583 static inline void skb_gro_incr_csum_unnecessary(struct sk_buff *skb) 2584 { 2585 if (NAPI_GRO_CB(skb)->csum_cnt > 0) { 2586 /* Consume a checksum from CHECKSUM_UNNECESSARY */ 2587 NAPI_GRO_CB(skb)->csum_cnt--; 2588 } else { 2589 /* Update skb for CHECKSUM_UNNECESSARY and csum_level when we 2590 * verified a new top level checksum or an encapsulated one 2591 * during GRO. This saves work if we fallback to normal path. 2592 */ 2593 __skb_incr_checksum_unnecessary(skb); 2594 } 2595 } 2596 2597 #define __skb_gro_checksum_validate(skb, proto, zero_okay, check, \ 2598 compute_pseudo) \ 2599 ({ \ 2600 __sum16 __ret = 0; \ 2601 if (__skb_gro_checksum_validate_needed(skb, zero_okay, check)) \ 2602 __ret = __skb_gro_checksum_validate_complete(skb, \ 2603 compute_pseudo(skb, proto)); \ 2604 if (!__ret) \ 2605 skb_gro_incr_csum_unnecessary(skb); \ 2606 __ret; \ 2607 }) 2608 2609 #define skb_gro_checksum_validate(skb, proto, compute_pseudo) \ 2610 __skb_gro_checksum_validate(skb, proto, false, 0, compute_pseudo) 2611 2612 #define skb_gro_checksum_validate_zero_check(skb, proto, check, \ 2613 compute_pseudo) \ 2614 __skb_gro_checksum_validate(skb, proto, true, check, compute_pseudo) 2615 2616 #define skb_gro_checksum_simple_validate(skb) \ 2617 __skb_gro_checksum_validate(skb, 0, false, 0, null_compute_pseudo) 2618 2619 static inline bool __skb_gro_checksum_convert_check(struct sk_buff *skb) 2620 { 2621 return (NAPI_GRO_CB(skb)->csum_cnt == 0 && 2622 !NAPI_GRO_CB(skb)->csum_valid); 2623 } 2624 2625 static inline void __skb_gro_checksum_convert(struct sk_buff *skb, 2626 __sum16 check, __wsum pseudo) 2627 { 2628 NAPI_GRO_CB(skb)->csum = ~pseudo; 2629 NAPI_GRO_CB(skb)->csum_valid = 1; 2630 } 2631 2632 #define skb_gro_checksum_try_convert(skb, proto, check, compute_pseudo) \ 2633 do { \ 2634 if (__skb_gro_checksum_convert_check(skb)) \ 2635 __skb_gro_checksum_convert(skb, check, \ 2636 compute_pseudo(skb, proto)); \ 2637 } while (0) 2638 2639 struct gro_remcsum { 2640 int offset; 2641 __wsum delta; 2642 }; 2643 2644 static inline void skb_gro_remcsum_init(struct gro_remcsum *grc) 2645 { 2646 grc->offset = 0; 2647 grc->delta = 0; 2648 } 2649 2650 static inline void *skb_gro_remcsum_process(struct sk_buff *skb, void *ptr, 2651 unsigned int off, size_t hdrlen, 2652 int start, int offset, 2653 struct gro_remcsum *grc, 2654 bool nopartial) 2655 { 2656 __wsum delta; 2657 size_t plen = hdrlen + max_t(size_t, offset + sizeof(u16), start); 2658 2659 BUG_ON(!NAPI_GRO_CB(skb)->csum_valid); 2660 2661 if (!nopartial) { 2662 NAPI_GRO_CB(skb)->gro_remcsum_start = off + hdrlen + start; 2663 return ptr; 2664 } 2665 2666 ptr = skb_gro_header_fast(skb, off); 2667 if (skb_gro_header_hard(skb, off + plen)) { 2668 ptr = skb_gro_header_slow(skb, off + plen, off); 2669 if (!ptr) 2670 return NULL; 2671 } 2672 2673 delta = remcsum_adjust(ptr + hdrlen, NAPI_GRO_CB(skb)->csum, 2674 start, offset); 2675 2676 /* Adjust skb->csum since we changed the packet */ 2677 NAPI_GRO_CB(skb)->csum = csum_add(NAPI_GRO_CB(skb)->csum, delta); 2678 2679 grc->offset = off + hdrlen + offset; 2680 grc->delta = delta; 2681 2682 return ptr; 2683 } 2684 2685 static inline void skb_gro_remcsum_cleanup(struct sk_buff *skb, 2686 struct gro_remcsum *grc) 2687 { 2688 void *ptr; 2689 size_t plen = grc->offset + sizeof(u16); 2690 2691 if (!grc->delta) 2692 return; 2693 2694 ptr = skb_gro_header_fast(skb, grc->offset); 2695 if (skb_gro_header_hard(skb, grc->offset + sizeof(u16))) { 2696 ptr = skb_gro_header_slow(skb, plen, grc->offset); 2697 if (!ptr) 2698 return; 2699 } 2700 2701 remcsum_unadjust((__sum16 *)ptr, grc->delta); 2702 } 2703 2704 #ifdef CONFIG_XFRM_OFFLOAD 2705 static inline void skb_gro_flush_final(struct sk_buff *skb, struct sk_buff **pp, int flush) 2706 { 2707 if (PTR_ERR(pp) != -EINPROGRESS) 2708 NAPI_GRO_CB(skb)->flush |= flush; 2709 } 2710 #else 2711 static inline void skb_gro_flush_final(struct sk_buff *skb, struct sk_buff **pp, int flush) 2712 { 2713 NAPI_GRO_CB(skb)->flush |= flush; 2714 } 2715 #endif 2716 2717 static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev, 2718 unsigned short type, 2719 const void *daddr, const void *saddr, 2720 unsigned int len) 2721 { 2722 if (!dev->header_ops || !dev->header_ops->create) 2723 return 0; 2724 2725 return dev->header_ops->create(skb, dev, type, daddr, saddr, len); 2726 } 2727 2728 static inline int dev_parse_header(const struct sk_buff *skb, 2729 unsigned char *haddr) 2730 { 2731 const struct net_device *dev = skb->dev; 2732 2733 if (!dev->header_ops || !dev->header_ops->parse) 2734 return 0; 2735 return dev->header_ops->parse(skb, haddr); 2736 } 2737 2738 /* ll_header must have at least hard_header_len allocated */ 2739 static inline bool dev_validate_header(const struct net_device *dev, 2740 char *ll_header, int len) 2741 { 2742 if (likely(len >= dev->hard_header_len)) 2743 return true; 2744 if (len < dev->min_header_len) 2745 return false; 2746 2747 if (capable(CAP_SYS_RAWIO)) { 2748 memset(ll_header + len, 0, dev->hard_header_len - len); 2749 return true; 2750 } 2751 2752 if (dev->header_ops && dev->header_ops->validate) 2753 return dev->header_ops->validate(ll_header, len); 2754 2755 return false; 2756 } 2757 2758 typedef int gifconf_func_t(struct net_device * dev, char __user * bufptr, int len); 2759 int register_gifconf(unsigned int family, gifconf_func_t *gifconf); 2760 static inline int unregister_gifconf(unsigned int family) 2761 { 2762 return register_gifconf(family, NULL); 2763 } 2764 2765 #ifdef CONFIG_NET_FLOW_LIMIT 2766 #define FLOW_LIMIT_HISTORY (1 << 7) /* must be ^2 and !overflow buckets */ 2767 struct sd_flow_limit { 2768 u64 count; 2769 unsigned int num_buckets; 2770 unsigned int history_head; 2771 u16 history[FLOW_LIMIT_HISTORY]; 2772 u8 buckets[]; 2773 }; 2774 2775 extern int netdev_flow_limit_table_len; 2776 #endif /* CONFIG_NET_FLOW_LIMIT */ 2777 2778 /* 2779 * Incoming packets are placed on per-CPU queues 2780 */ 2781 struct softnet_data { 2782 struct list_head poll_list; 2783 struct sk_buff_head process_queue; 2784 2785 /* stats */ 2786 unsigned int processed; 2787 unsigned int time_squeeze; 2788 unsigned int received_rps; 2789 #ifdef CONFIG_RPS 2790 struct softnet_data *rps_ipi_list; 2791 #endif 2792 #ifdef CONFIG_NET_FLOW_LIMIT 2793 struct sd_flow_limit __rcu *flow_limit; 2794 #endif 2795 struct Qdisc *output_queue; 2796 struct Qdisc **output_queue_tailp; 2797 struct sk_buff *completion_queue; 2798 #ifdef CONFIG_XFRM_OFFLOAD 2799 struct sk_buff_head xfrm_backlog; 2800 #endif 2801 #ifdef CONFIG_RPS 2802 /* input_queue_head should be written by cpu owning this struct, 2803 * and only read by other cpus. Worth using a cache line. 2804 */ 2805 unsigned int input_queue_head ____cacheline_aligned_in_smp; 2806 2807 /* Elements below can be accessed between CPUs for RPS/RFS */ 2808 call_single_data_t csd ____cacheline_aligned_in_smp; 2809 struct softnet_data *rps_ipi_next; 2810 unsigned int cpu; 2811 unsigned int input_queue_tail; 2812 #endif 2813 unsigned int dropped; 2814 struct sk_buff_head input_pkt_queue; 2815 struct napi_struct backlog; 2816 2817 }; 2818 2819 static inline void input_queue_head_incr(struct softnet_data *sd) 2820 { 2821 #ifdef CONFIG_RPS 2822 sd->input_queue_head++; 2823 #endif 2824 } 2825 2826 static inline void input_queue_tail_incr_save(struct softnet_data *sd, 2827 unsigned int *qtail) 2828 { 2829 #ifdef CONFIG_RPS 2830 *qtail = ++sd->input_queue_tail; 2831 #endif 2832 } 2833 2834 DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data); 2835 2836 void __netif_schedule(struct Qdisc *q); 2837 void netif_schedule_queue(struct netdev_queue *txq); 2838 2839 static inline void netif_tx_schedule_all(struct net_device *dev) 2840 { 2841 unsigned int i; 2842 2843 for (i = 0; i < dev->num_tx_queues; i++) 2844 netif_schedule_queue(netdev_get_tx_queue(dev, i)); 2845 } 2846 2847 static __always_inline void netif_tx_start_queue(struct netdev_queue *dev_queue) 2848 { 2849 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state); 2850 } 2851 2852 /** 2853 * netif_start_queue - allow transmit 2854 * @dev: network device 2855 * 2856 * Allow upper layers to call the device hard_start_xmit routine. 2857 */ 2858 static inline void netif_start_queue(struct net_device *dev) 2859 { 2860 netif_tx_start_queue(netdev_get_tx_queue(dev, 0)); 2861 } 2862 2863 static inline void netif_tx_start_all_queues(struct net_device *dev) 2864 { 2865 unsigned int i; 2866 2867 for (i = 0; i < dev->num_tx_queues; i++) { 2868 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 2869 netif_tx_start_queue(txq); 2870 } 2871 } 2872 2873 void netif_tx_wake_queue(struct netdev_queue *dev_queue); 2874 2875 /** 2876 * netif_wake_queue - restart transmit 2877 * @dev: network device 2878 * 2879 * Allow upper layers to call the device hard_start_xmit routine. 2880 * Used for flow control when transmit resources are available. 2881 */ 2882 static inline void netif_wake_queue(struct net_device *dev) 2883 { 2884 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0)); 2885 } 2886 2887 static inline void netif_tx_wake_all_queues(struct net_device *dev) 2888 { 2889 unsigned int i; 2890 2891 for (i = 0; i < dev->num_tx_queues; i++) { 2892 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 2893 netif_tx_wake_queue(txq); 2894 } 2895 } 2896 2897 static __always_inline void netif_tx_stop_queue(struct netdev_queue *dev_queue) 2898 { 2899 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state); 2900 } 2901 2902 /** 2903 * netif_stop_queue - stop transmitted packets 2904 * @dev: network device 2905 * 2906 * Stop upper layers calling the device hard_start_xmit routine. 2907 * Used for flow control when transmit resources are unavailable. 2908 */ 2909 static inline void netif_stop_queue(struct net_device *dev) 2910 { 2911 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0)); 2912 } 2913 2914 void netif_tx_stop_all_queues(struct net_device *dev); 2915 2916 static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue) 2917 { 2918 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state); 2919 } 2920 2921 /** 2922 * netif_queue_stopped - test if transmit queue is flowblocked 2923 * @dev: network device 2924 * 2925 * Test if transmit queue on device is currently unable to send. 2926 */ 2927 static inline bool netif_queue_stopped(const struct net_device *dev) 2928 { 2929 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0)); 2930 } 2931 2932 static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue) 2933 { 2934 return dev_queue->state & QUEUE_STATE_ANY_XOFF; 2935 } 2936 2937 static inline bool 2938 netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue) 2939 { 2940 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN; 2941 } 2942 2943 static inline bool 2944 netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue) 2945 { 2946 return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN; 2947 } 2948 2949 /** 2950 * netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write 2951 * @dev_queue: pointer to transmit queue 2952 * 2953 * BQL enabled drivers might use this helper in their ndo_start_xmit(), 2954 * to give appropriate hint to the CPU. 2955 */ 2956 static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue) 2957 { 2958 #ifdef CONFIG_BQL 2959 prefetchw(&dev_queue->dql.num_queued); 2960 #endif 2961 } 2962 2963 /** 2964 * netdev_txq_bql_complete_prefetchw - prefetch bql data for write 2965 * @dev_queue: pointer to transmit queue 2966 * 2967 * BQL enabled drivers might use this helper in their TX completion path, 2968 * to give appropriate hint to the CPU. 2969 */ 2970 static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue) 2971 { 2972 #ifdef CONFIG_BQL 2973 prefetchw(&dev_queue->dql.limit); 2974 #endif 2975 } 2976 2977 static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue, 2978 unsigned int bytes) 2979 { 2980 #ifdef CONFIG_BQL 2981 dql_queued(&dev_queue->dql, bytes); 2982 2983 if (likely(dql_avail(&dev_queue->dql) >= 0)) 2984 return; 2985 2986 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state); 2987 2988 /* 2989 * The XOFF flag must be set before checking the dql_avail below, 2990 * because in netdev_tx_completed_queue we update the dql_completed 2991 * before checking the XOFF flag. 2992 */ 2993 smp_mb(); 2994 2995 /* check again in case another CPU has just made room avail */ 2996 if (unlikely(dql_avail(&dev_queue->dql) >= 0)) 2997 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state); 2998 #endif 2999 } 3000 3001 /** 3002 * netdev_sent_queue - report the number of bytes queued to hardware 3003 * @dev: network device 3004 * @bytes: number of bytes queued to the hardware device queue 3005 * 3006 * Report the number of bytes queued for sending/completion to the network 3007 * device hardware queue. @bytes should be a good approximation and should 3008 * exactly match netdev_completed_queue() @bytes 3009 */ 3010 static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes) 3011 { 3012 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes); 3013 } 3014 3015 static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue, 3016 unsigned int pkts, unsigned int bytes) 3017 { 3018 #ifdef CONFIG_BQL 3019 if (unlikely(!bytes)) 3020 return; 3021 3022 dql_completed(&dev_queue->dql, bytes); 3023 3024 /* 3025 * Without the memory barrier there is a small possiblity that 3026 * netdev_tx_sent_queue will miss the update and cause the queue to 3027 * be stopped forever 3028 */ 3029 smp_mb(); 3030 3031 if (dql_avail(&dev_queue->dql) < 0) 3032 return; 3033 3034 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state)) 3035 netif_schedule_queue(dev_queue); 3036 #endif 3037 } 3038 3039 /** 3040 * netdev_completed_queue - report bytes and packets completed by device 3041 * @dev: network device 3042 * @pkts: actual number of packets sent over the medium 3043 * @bytes: actual number of bytes sent over the medium 3044 * 3045 * Report the number of bytes and packets transmitted by the network device 3046 * hardware queue over the physical medium, @bytes must exactly match the 3047 * @bytes amount passed to netdev_sent_queue() 3048 */ 3049 static inline void netdev_completed_queue(struct net_device *dev, 3050 unsigned int pkts, unsigned int bytes) 3051 { 3052 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes); 3053 } 3054 3055 static inline void netdev_tx_reset_queue(struct netdev_queue *q) 3056 { 3057 #ifdef CONFIG_BQL 3058 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state); 3059 dql_reset(&q->dql); 3060 #endif 3061 } 3062 3063 /** 3064 * netdev_reset_queue - reset the packets and bytes count of a network device 3065 * @dev_queue: network device 3066 * 3067 * Reset the bytes and packet count of a network device and clear the 3068 * software flow control OFF bit for this network device 3069 */ 3070 static inline void netdev_reset_queue(struct net_device *dev_queue) 3071 { 3072 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0)); 3073 } 3074 3075 /** 3076 * netdev_cap_txqueue - check if selected tx queue exceeds device queues 3077 * @dev: network device 3078 * @queue_index: given tx queue index 3079 * 3080 * Returns 0 if given tx queue index >= number of device tx queues, 3081 * otherwise returns the originally passed tx queue index. 3082 */ 3083 static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index) 3084 { 3085 if (unlikely(queue_index >= dev->real_num_tx_queues)) { 3086 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n", 3087 dev->name, queue_index, 3088 dev->real_num_tx_queues); 3089 return 0; 3090 } 3091 3092 return queue_index; 3093 } 3094 3095 /** 3096 * netif_running - test if up 3097 * @dev: network device 3098 * 3099 * Test if the device has been brought up. 3100 */ 3101 static inline bool netif_running(const struct net_device *dev) 3102 { 3103 return test_bit(__LINK_STATE_START, &dev->state); 3104 } 3105 3106 /* 3107 * Routines to manage the subqueues on a device. We only need start, 3108 * stop, and a check if it's stopped. All other device management is 3109 * done at the overall netdevice level. 3110 * Also test the device if we're multiqueue. 3111 */ 3112 3113 /** 3114 * netif_start_subqueue - allow sending packets on subqueue 3115 * @dev: network device 3116 * @queue_index: sub queue index 3117 * 3118 * Start individual transmit queue of a device with multiple transmit queues. 3119 */ 3120 static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index) 3121 { 3122 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 3123 3124 netif_tx_start_queue(txq); 3125 } 3126 3127 /** 3128 * netif_stop_subqueue - stop sending packets on subqueue 3129 * @dev: network device 3130 * @queue_index: sub queue index 3131 * 3132 * Stop individual transmit queue of a device with multiple transmit queues. 3133 */ 3134 static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index) 3135 { 3136 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 3137 netif_tx_stop_queue(txq); 3138 } 3139 3140 /** 3141 * netif_subqueue_stopped - test status of subqueue 3142 * @dev: network device 3143 * @queue_index: sub queue index 3144 * 3145 * Check individual transmit queue of a device with multiple transmit queues. 3146 */ 3147 static inline bool __netif_subqueue_stopped(const struct net_device *dev, 3148 u16 queue_index) 3149 { 3150 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 3151 3152 return netif_tx_queue_stopped(txq); 3153 } 3154 3155 static inline bool netif_subqueue_stopped(const struct net_device *dev, 3156 struct sk_buff *skb) 3157 { 3158 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb)); 3159 } 3160 3161 /** 3162 * netif_wake_subqueue - allow sending packets on subqueue 3163 * @dev: network device 3164 * @queue_index: sub queue index 3165 * 3166 * Resume individual transmit queue of a device with multiple transmit queues. 3167 */ 3168 static inline void netif_wake_subqueue(struct net_device *dev, u16 queue_index) 3169 { 3170 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 3171 3172 netif_tx_wake_queue(txq); 3173 } 3174 3175 #ifdef CONFIG_XPS 3176 int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask, 3177 u16 index); 3178 #else 3179 static inline int netif_set_xps_queue(struct net_device *dev, 3180 const struct cpumask *mask, 3181 u16 index) 3182 { 3183 return 0; 3184 } 3185 #endif 3186 3187 u16 __skb_tx_hash(const struct net_device *dev, struct sk_buff *skb, 3188 unsigned int num_tx_queues); 3189 3190 /* 3191 * Returns a Tx hash for the given packet when dev->real_num_tx_queues is used 3192 * as a distribution range limit for the returned value. 3193 */ 3194 static inline u16 skb_tx_hash(const struct net_device *dev, 3195 struct sk_buff *skb) 3196 { 3197 return __skb_tx_hash(dev, skb, dev->real_num_tx_queues); 3198 } 3199 3200 /** 3201 * netif_is_multiqueue - test if device has multiple transmit queues 3202 * @dev: network device 3203 * 3204 * Check if device has multiple transmit queues 3205 */ 3206 static inline bool netif_is_multiqueue(const struct net_device *dev) 3207 { 3208 return dev->num_tx_queues > 1; 3209 } 3210 3211 int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq); 3212 3213 #ifdef CONFIG_SYSFS 3214 int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq); 3215 #else 3216 static inline int netif_set_real_num_rx_queues(struct net_device *dev, 3217 unsigned int rxq) 3218 { 3219 return 0; 3220 } 3221 #endif 3222 3223 #ifdef CONFIG_SYSFS 3224 static inline unsigned int get_netdev_rx_queue_index( 3225 struct netdev_rx_queue *queue) 3226 { 3227 struct net_device *dev = queue->dev; 3228 int index = queue - dev->_rx; 3229 3230 BUG_ON(index >= dev->num_rx_queues); 3231 return index; 3232 } 3233 #endif 3234 3235 #define DEFAULT_MAX_NUM_RSS_QUEUES (8) 3236 int netif_get_num_default_rss_queues(void); 3237 3238 enum skb_free_reason { 3239 SKB_REASON_CONSUMED, 3240 SKB_REASON_DROPPED, 3241 }; 3242 3243 void __dev_kfree_skb_irq(struct sk_buff *skb, enum skb_free_reason reason); 3244 void __dev_kfree_skb_any(struct sk_buff *skb, enum skb_free_reason reason); 3245 3246 /* 3247 * It is not allowed to call kfree_skb() or consume_skb() from hardware 3248 * interrupt context or with hardware interrupts being disabled. 3249 * (in_irq() || irqs_disabled()) 3250 * 3251 * We provide four helpers that can be used in following contexts : 3252 * 3253 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context, 3254 * replacing kfree_skb(skb) 3255 * 3256 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context. 3257 * Typically used in place of consume_skb(skb) in TX completion path 3258 * 3259 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context, 3260 * replacing kfree_skb(skb) 3261 * 3262 * dev_consume_skb_any(skb) when caller doesn't know its current irq context, 3263 * and consumed a packet. Used in place of consume_skb(skb) 3264 */ 3265 static inline void dev_kfree_skb_irq(struct sk_buff *skb) 3266 { 3267 __dev_kfree_skb_irq(skb, SKB_REASON_DROPPED); 3268 } 3269 3270 static inline void dev_consume_skb_irq(struct sk_buff *skb) 3271 { 3272 __dev_kfree_skb_irq(skb, SKB_REASON_CONSUMED); 3273 } 3274 3275 static inline void dev_kfree_skb_any(struct sk_buff *skb) 3276 { 3277 __dev_kfree_skb_any(skb, SKB_REASON_DROPPED); 3278 } 3279 3280 static inline void dev_consume_skb_any(struct sk_buff *skb) 3281 { 3282 __dev_kfree_skb_any(skb, SKB_REASON_CONSUMED); 3283 } 3284 3285 void generic_xdp_tx(struct sk_buff *skb, struct bpf_prog *xdp_prog); 3286 int do_xdp_generic(struct bpf_prog *xdp_prog, struct sk_buff *skb); 3287 int netif_rx(struct sk_buff *skb); 3288 int netif_rx_ni(struct sk_buff *skb); 3289 int netif_receive_skb(struct sk_buff *skb); 3290 int netif_receive_skb_core(struct sk_buff *skb); 3291 gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb); 3292 void napi_gro_flush(struct napi_struct *napi, bool flush_old); 3293 struct sk_buff *napi_get_frags(struct napi_struct *napi); 3294 gro_result_t napi_gro_frags(struct napi_struct *napi); 3295 struct packet_offload *gro_find_receive_by_type(__be16 type); 3296 struct packet_offload *gro_find_complete_by_type(__be16 type); 3297 3298 static inline void napi_free_frags(struct napi_struct *napi) 3299 { 3300 kfree_skb(napi->skb); 3301 napi->skb = NULL; 3302 } 3303 3304 bool netdev_is_rx_handler_busy(struct net_device *dev); 3305 int netdev_rx_handler_register(struct net_device *dev, 3306 rx_handler_func_t *rx_handler, 3307 void *rx_handler_data); 3308 void netdev_rx_handler_unregister(struct net_device *dev); 3309 3310 bool dev_valid_name(const char *name); 3311 int dev_ioctl(struct net *net, unsigned int cmd, void __user *); 3312 int dev_ethtool(struct net *net, struct ifreq *); 3313 unsigned int dev_get_flags(const struct net_device *); 3314 int __dev_change_flags(struct net_device *, unsigned int flags); 3315 int dev_change_flags(struct net_device *, unsigned int); 3316 void __dev_notify_flags(struct net_device *, unsigned int old_flags, 3317 unsigned int gchanges); 3318 int dev_change_name(struct net_device *, const char *); 3319 int dev_set_alias(struct net_device *, const char *, size_t); 3320 int dev_get_alias(const struct net_device *, char *, size_t); 3321 int dev_change_net_namespace(struct net_device *, struct net *, const char *); 3322 int __dev_set_mtu(struct net_device *, int); 3323 int dev_set_mtu(struct net_device *, int); 3324 void dev_set_group(struct net_device *, int); 3325 int dev_set_mac_address(struct net_device *, struct sockaddr *); 3326 int dev_change_carrier(struct net_device *, bool new_carrier); 3327 int dev_get_phys_port_id(struct net_device *dev, 3328 struct netdev_phys_item_id *ppid); 3329 int dev_get_phys_port_name(struct net_device *dev, 3330 char *name, size_t len); 3331 int dev_change_proto_down(struct net_device *dev, bool proto_down); 3332 struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev, bool *again); 3333 struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev, 3334 struct netdev_queue *txq, int *ret); 3335 3336 typedef int (*bpf_op_t)(struct net_device *dev, struct netdev_bpf *bpf); 3337 int dev_change_xdp_fd(struct net_device *dev, struct netlink_ext_ack *extack, 3338 int fd, u32 flags); 3339 void __dev_xdp_query(struct net_device *dev, bpf_op_t xdp_op, 3340 struct netdev_bpf *xdp); 3341 3342 int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb); 3343 int dev_forward_skb(struct net_device *dev, struct sk_buff *skb); 3344 bool is_skb_forwardable(const struct net_device *dev, 3345 const struct sk_buff *skb); 3346 3347 static __always_inline int ____dev_forward_skb(struct net_device *dev, 3348 struct sk_buff *skb) 3349 { 3350 if (skb_orphan_frags(skb, GFP_ATOMIC) || 3351 unlikely(!is_skb_forwardable(dev, skb))) { 3352 atomic_long_inc(&dev->rx_dropped); 3353 kfree_skb(skb); 3354 return NET_RX_DROP; 3355 } 3356 3357 skb_scrub_packet(skb, true); 3358 skb->priority = 0; 3359 return 0; 3360 } 3361 3362 void dev_queue_xmit_nit(struct sk_buff *skb, struct net_device *dev); 3363 3364 extern int netdev_budget; 3365 extern unsigned int netdev_budget_usecs; 3366 3367 /* Called by rtnetlink.c:rtnl_unlock() */ 3368 void netdev_run_todo(void); 3369 3370 /** 3371 * dev_put - release reference to device 3372 * @dev: network device 3373 * 3374 * Release reference to device to allow it to be freed. 3375 */ 3376 static inline void dev_put(struct net_device *dev) 3377 { 3378 this_cpu_dec(*dev->pcpu_refcnt); 3379 } 3380 3381 /** 3382 * dev_hold - get reference to device 3383 * @dev: network device 3384 * 3385 * Hold reference to device to keep it from being freed. 3386 */ 3387 static inline void dev_hold(struct net_device *dev) 3388 { 3389 this_cpu_inc(*dev->pcpu_refcnt); 3390 } 3391 3392 /* Carrier loss detection, dial on demand. The functions netif_carrier_on 3393 * and _off may be called from IRQ context, but it is caller 3394 * who is responsible for serialization of these calls. 3395 * 3396 * The name carrier is inappropriate, these functions should really be 3397 * called netif_lowerlayer_*() because they represent the state of any 3398 * kind of lower layer not just hardware media. 3399 */ 3400 3401 void linkwatch_init_dev(struct net_device *dev); 3402 void linkwatch_fire_event(struct net_device *dev); 3403 void linkwatch_forget_dev(struct net_device *dev); 3404 3405 /** 3406 * netif_carrier_ok - test if carrier present 3407 * @dev: network device 3408 * 3409 * Check if carrier is present on device 3410 */ 3411 static inline bool netif_carrier_ok(const struct net_device *dev) 3412 { 3413 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state); 3414 } 3415 3416 unsigned long dev_trans_start(struct net_device *dev); 3417 3418 void __netdev_watchdog_up(struct net_device *dev); 3419 3420 void netif_carrier_on(struct net_device *dev); 3421 3422 void netif_carrier_off(struct net_device *dev); 3423 3424 /** 3425 * netif_dormant_on - mark device as dormant. 3426 * @dev: network device 3427 * 3428 * Mark device as dormant (as per RFC2863). 3429 * 3430 * The dormant state indicates that the relevant interface is not 3431 * actually in a condition to pass packets (i.e., it is not 'up') but is 3432 * in a "pending" state, waiting for some external event. For "on- 3433 * demand" interfaces, this new state identifies the situation where the 3434 * interface is waiting for events to place it in the up state. 3435 */ 3436 static inline void netif_dormant_on(struct net_device *dev) 3437 { 3438 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state)) 3439 linkwatch_fire_event(dev); 3440 } 3441 3442 /** 3443 * netif_dormant_off - set device as not dormant. 3444 * @dev: network device 3445 * 3446 * Device is not in dormant state. 3447 */ 3448 static inline void netif_dormant_off(struct net_device *dev) 3449 { 3450 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state)) 3451 linkwatch_fire_event(dev); 3452 } 3453 3454 /** 3455 * netif_dormant - test if device is dormant 3456 * @dev: network device 3457 * 3458 * Check if device is dormant. 3459 */ 3460 static inline bool netif_dormant(const struct net_device *dev) 3461 { 3462 return test_bit(__LINK_STATE_DORMANT, &dev->state); 3463 } 3464 3465 3466 /** 3467 * netif_oper_up - test if device is operational 3468 * @dev: network device 3469 * 3470 * Check if carrier is operational 3471 */ 3472 static inline bool netif_oper_up(const struct net_device *dev) 3473 { 3474 return (dev->operstate == IF_OPER_UP || 3475 dev->operstate == IF_OPER_UNKNOWN /* backward compat */); 3476 } 3477 3478 /** 3479 * netif_device_present - is device available or removed 3480 * @dev: network device 3481 * 3482 * Check if device has not been removed from system. 3483 */ 3484 static inline bool netif_device_present(struct net_device *dev) 3485 { 3486 return test_bit(__LINK_STATE_PRESENT, &dev->state); 3487 } 3488 3489 void netif_device_detach(struct net_device *dev); 3490 3491 void netif_device_attach(struct net_device *dev); 3492 3493 /* 3494 * Network interface message level settings 3495 */ 3496 3497 enum { 3498 NETIF_MSG_DRV = 0x0001, 3499 NETIF_MSG_PROBE = 0x0002, 3500 NETIF_MSG_LINK = 0x0004, 3501 NETIF_MSG_TIMER = 0x0008, 3502 NETIF_MSG_IFDOWN = 0x0010, 3503 NETIF_MSG_IFUP = 0x0020, 3504 NETIF_MSG_RX_ERR = 0x0040, 3505 NETIF_MSG_TX_ERR = 0x0080, 3506 NETIF_MSG_TX_QUEUED = 0x0100, 3507 NETIF_MSG_INTR = 0x0200, 3508 NETIF_MSG_TX_DONE = 0x0400, 3509 NETIF_MSG_RX_STATUS = 0x0800, 3510 NETIF_MSG_PKTDATA = 0x1000, 3511 NETIF_MSG_HW = 0x2000, 3512 NETIF_MSG_WOL = 0x4000, 3513 }; 3514 3515 #define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV) 3516 #define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE) 3517 #define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK) 3518 #define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER) 3519 #define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN) 3520 #define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP) 3521 #define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR) 3522 #define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR) 3523 #define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED) 3524 #define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR) 3525 #define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE) 3526 #define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS) 3527 #define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA) 3528 #define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW) 3529 #define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL) 3530 3531 static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits) 3532 { 3533 /* use default */ 3534 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8)) 3535 return default_msg_enable_bits; 3536 if (debug_value == 0) /* no output */ 3537 return 0; 3538 /* set low N bits */ 3539 return (1 << debug_value) - 1; 3540 } 3541 3542 static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu) 3543 { 3544 spin_lock(&txq->_xmit_lock); 3545 txq->xmit_lock_owner = cpu; 3546 } 3547 3548 static inline bool __netif_tx_acquire(struct netdev_queue *txq) 3549 { 3550 __acquire(&txq->_xmit_lock); 3551 return true; 3552 } 3553 3554 static inline void __netif_tx_release(struct netdev_queue *txq) 3555 { 3556 __release(&txq->_xmit_lock); 3557 } 3558 3559 static inline void __netif_tx_lock_bh(struct netdev_queue *txq) 3560 { 3561 spin_lock_bh(&txq->_xmit_lock); 3562 txq->xmit_lock_owner = smp_processor_id(); 3563 } 3564 3565 static inline bool __netif_tx_trylock(struct netdev_queue *txq) 3566 { 3567 bool ok = spin_trylock(&txq->_xmit_lock); 3568 if (likely(ok)) 3569 txq->xmit_lock_owner = smp_processor_id(); 3570 return ok; 3571 } 3572 3573 static inline void __netif_tx_unlock(struct netdev_queue *txq) 3574 { 3575 txq->xmit_lock_owner = -1; 3576 spin_unlock(&txq->_xmit_lock); 3577 } 3578 3579 static inline void __netif_tx_unlock_bh(struct netdev_queue *txq) 3580 { 3581 txq->xmit_lock_owner = -1; 3582 spin_unlock_bh(&txq->_xmit_lock); 3583 } 3584 3585 static inline void txq_trans_update(struct netdev_queue *txq) 3586 { 3587 if (txq->xmit_lock_owner != -1) 3588 txq->trans_start = jiffies; 3589 } 3590 3591 /* legacy drivers only, netdev_start_xmit() sets txq->trans_start */ 3592 static inline void netif_trans_update(struct net_device *dev) 3593 { 3594 struct netdev_queue *txq = netdev_get_tx_queue(dev, 0); 3595 3596 if (txq->trans_start != jiffies) 3597 txq->trans_start = jiffies; 3598 } 3599 3600 /** 3601 * netif_tx_lock - grab network device transmit lock 3602 * @dev: network device 3603 * 3604 * Get network device transmit lock 3605 */ 3606 static inline void netif_tx_lock(struct net_device *dev) 3607 { 3608 unsigned int i; 3609 int cpu; 3610 3611 spin_lock(&dev->tx_global_lock); 3612 cpu = smp_processor_id(); 3613 for (i = 0; i < dev->num_tx_queues; i++) { 3614 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 3615 3616 /* We are the only thread of execution doing a 3617 * freeze, but we have to grab the _xmit_lock in 3618 * order to synchronize with threads which are in 3619 * the ->hard_start_xmit() handler and already 3620 * checked the frozen bit. 3621 */ 3622 __netif_tx_lock(txq, cpu); 3623 set_bit(__QUEUE_STATE_FROZEN, &txq->state); 3624 __netif_tx_unlock(txq); 3625 } 3626 } 3627 3628 static inline void netif_tx_lock_bh(struct net_device *dev) 3629 { 3630 local_bh_disable(); 3631 netif_tx_lock(dev); 3632 } 3633 3634 static inline void netif_tx_unlock(struct net_device *dev) 3635 { 3636 unsigned int i; 3637 3638 for (i = 0; i < dev->num_tx_queues; i++) { 3639 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 3640 3641 /* No need to grab the _xmit_lock here. If the 3642 * queue is not stopped for another reason, we 3643 * force a schedule. 3644 */ 3645 clear_bit(__QUEUE_STATE_FROZEN, &txq->state); 3646 netif_schedule_queue(txq); 3647 } 3648 spin_unlock(&dev->tx_global_lock); 3649 } 3650 3651 static inline void netif_tx_unlock_bh(struct net_device *dev) 3652 { 3653 netif_tx_unlock(dev); 3654 local_bh_enable(); 3655 } 3656 3657 #define HARD_TX_LOCK(dev, txq, cpu) { \ 3658 if ((dev->features & NETIF_F_LLTX) == 0) { \ 3659 __netif_tx_lock(txq, cpu); \ 3660 } else { \ 3661 __netif_tx_acquire(txq); \ 3662 } \ 3663 } 3664 3665 #define HARD_TX_TRYLOCK(dev, txq) \ 3666 (((dev->features & NETIF_F_LLTX) == 0) ? \ 3667 __netif_tx_trylock(txq) : \ 3668 __netif_tx_acquire(txq)) 3669 3670 #define HARD_TX_UNLOCK(dev, txq) { \ 3671 if ((dev->features & NETIF_F_LLTX) == 0) { \ 3672 __netif_tx_unlock(txq); \ 3673 } else { \ 3674 __netif_tx_release(txq); \ 3675 } \ 3676 } 3677 3678 static inline void netif_tx_disable(struct net_device *dev) 3679 { 3680 unsigned int i; 3681 int cpu; 3682 3683 local_bh_disable(); 3684 cpu = smp_processor_id(); 3685 for (i = 0; i < dev->num_tx_queues; i++) { 3686 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 3687 3688 __netif_tx_lock(txq, cpu); 3689 netif_tx_stop_queue(txq); 3690 __netif_tx_unlock(txq); 3691 } 3692 local_bh_enable(); 3693 } 3694 3695 static inline void netif_addr_lock(struct net_device *dev) 3696 { 3697 spin_lock(&dev->addr_list_lock); 3698 } 3699 3700 static inline void netif_addr_lock_nested(struct net_device *dev) 3701 { 3702 int subclass = SINGLE_DEPTH_NESTING; 3703 3704 if (dev->netdev_ops->ndo_get_lock_subclass) 3705 subclass = dev->netdev_ops->ndo_get_lock_subclass(dev); 3706 3707 spin_lock_nested(&dev->addr_list_lock, subclass); 3708 } 3709 3710 static inline void netif_addr_lock_bh(struct net_device *dev) 3711 { 3712 spin_lock_bh(&dev->addr_list_lock); 3713 } 3714 3715 static inline void netif_addr_unlock(struct net_device *dev) 3716 { 3717 spin_unlock(&dev->addr_list_lock); 3718 } 3719 3720 static inline void netif_addr_unlock_bh(struct net_device *dev) 3721 { 3722 spin_unlock_bh(&dev->addr_list_lock); 3723 } 3724 3725 /* 3726 * dev_addrs walker. Should be used only for read access. Call with 3727 * rcu_read_lock held. 3728 */ 3729 #define for_each_dev_addr(dev, ha) \ 3730 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list) 3731 3732 /* These functions live elsewhere (drivers/net/net_init.c, but related) */ 3733 3734 void ether_setup(struct net_device *dev); 3735 3736 /* Support for loadable net-drivers */ 3737 struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name, 3738 unsigned char name_assign_type, 3739 void (*setup)(struct net_device *), 3740 unsigned int txqs, unsigned int rxqs); 3741 int dev_get_valid_name(struct net *net, struct net_device *dev, 3742 const char *name); 3743 3744 #define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \ 3745 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1) 3746 3747 #define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \ 3748 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \ 3749 count) 3750 3751 int register_netdev(struct net_device *dev); 3752 void unregister_netdev(struct net_device *dev); 3753 3754 /* General hardware address lists handling functions */ 3755 int __hw_addr_sync(struct netdev_hw_addr_list *to_list, 3756 struct netdev_hw_addr_list *from_list, int addr_len); 3757 void __hw_addr_unsync(struct netdev_hw_addr_list *to_list, 3758 struct netdev_hw_addr_list *from_list, int addr_len); 3759 int __hw_addr_sync_dev(struct netdev_hw_addr_list *list, 3760 struct net_device *dev, 3761 int (*sync)(struct net_device *, const unsigned char *), 3762 int (*unsync)(struct net_device *, 3763 const unsigned char *)); 3764 void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list, 3765 struct net_device *dev, 3766 int (*unsync)(struct net_device *, 3767 const unsigned char *)); 3768 void __hw_addr_init(struct netdev_hw_addr_list *list); 3769 3770 /* Functions used for device addresses handling */ 3771 int dev_addr_add(struct net_device *dev, const unsigned char *addr, 3772 unsigned char addr_type); 3773 int dev_addr_del(struct net_device *dev, const unsigned char *addr, 3774 unsigned char addr_type); 3775 void dev_addr_flush(struct net_device *dev); 3776 int dev_addr_init(struct net_device *dev); 3777 3778 /* Functions used for unicast addresses handling */ 3779 int dev_uc_add(struct net_device *dev, const unsigned char *addr); 3780 int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr); 3781 int dev_uc_del(struct net_device *dev, const unsigned char *addr); 3782 int dev_uc_sync(struct net_device *to, struct net_device *from); 3783 int dev_uc_sync_multiple(struct net_device *to, struct net_device *from); 3784 void dev_uc_unsync(struct net_device *to, struct net_device *from); 3785 void dev_uc_flush(struct net_device *dev); 3786 void dev_uc_init(struct net_device *dev); 3787 3788 /** 3789 * __dev_uc_sync - Synchonize device's unicast list 3790 * @dev: device to sync 3791 * @sync: function to call if address should be added 3792 * @unsync: function to call if address should be removed 3793 * 3794 * Add newly added addresses to the interface, and release 3795 * addresses that have been deleted. 3796 */ 3797 static inline int __dev_uc_sync(struct net_device *dev, 3798 int (*sync)(struct net_device *, 3799 const unsigned char *), 3800 int (*unsync)(struct net_device *, 3801 const unsigned char *)) 3802 { 3803 return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync); 3804 } 3805 3806 /** 3807 * __dev_uc_unsync - Remove synchronized addresses from device 3808 * @dev: device to sync 3809 * @unsync: function to call if address should be removed 3810 * 3811 * Remove all addresses that were added to the device by dev_uc_sync(). 3812 */ 3813 static inline void __dev_uc_unsync(struct net_device *dev, 3814 int (*unsync)(struct net_device *, 3815 const unsigned char *)) 3816 { 3817 __hw_addr_unsync_dev(&dev->uc, dev, unsync); 3818 } 3819 3820 /* Functions used for multicast addresses handling */ 3821 int dev_mc_add(struct net_device *dev, const unsigned char *addr); 3822 int dev_mc_add_global(struct net_device *dev, const unsigned char *addr); 3823 int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr); 3824 int dev_mc_del(struct net_device *dev, const unsigned char *addr); 3825 int dev_mc_del_global(struct net_device *dev, const unsigned char *addr); 3826 int dev_mc_sync(struct net_device *to, struct net_device *from); 3827 int dev_mc_sync_multiple(struct net_device *to, struct net_device *from); 3828 void dev_mc_unsync(struct net_device *to, struct net_device *from); 3829 void dev_mc_flush(struct net_device *dev); 3830 void dev_mc_init(struct net_device *dev); 3831 3832 /** 3833 * __dev_mc_sync - Synchonize device's multicast list 3834 * @dev: device to sync 3835 * @sync: function to call if address should be added 3836 * @unsync: function to call if address should be removed 3837 * 3838 * Add newly added addresses to the interface, and release 3839 * addresses that have been deleted. 3840 */ 3841 static inline int __dev_mc_sync(struct net_device *dev, 3842 int (*sync)(struct net_device *, 3843 const unsigned char *), 3844 int (*unsync)(struct net_device *, 3845 const unsigned char *)) 3846 { 3847 return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync); 3848 } 3849 3850 /** 3851 * __dev_mc_unsync - Remove synchronized addresses from device 3852 * @dev: device to sync 3853 * @unsync: function to call if address should be removed 3854 * 3855 * Remove all addresses that were added to the device by dev_mc_sync(). 3856 */ 3857 static inline void __dev_mc_unsync(struct net_device *dev, 3858 int (*unsync)(struct net_device *, 3859 const unsigned char *)) 3860 { 3861 __hw_addr_unsync_dev(&dev->mc, dev, unsync); 3862 } 3863 3864 /* Functions used for secondary unicast and multicast support */ 3865 void dev_set_rx_mode(struct net_device *dev); 3866 void __dev_set_rx_mode(struct net_device *dev); 3867 int dev_set_promiscuity(struct net_device *dev, int inc); 3868 int dev_set_allmulti(struct net_device *dev, int inc); 3869 void netdev_state_change(struct net_device *dev); 3870 void netdev_notify_peers(struct net_device *dev); 3871 void netdev_features_change(struct net_device *dev); 3872 /* Load a device via the kmod */ 3873 void dev_load(struct net *net, const char *name); 3874 struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev, 3875 struct rtnl_link_stats64 *storage); 3876 void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64, 3877 const struct net_device_stats *netdev_stats); 3878 3879 extern int netdev_max_backlog; 3880 extern int netdev_tstamp_prequeue; 3881 extern int weight_p; 3882 extern int dev_weight_rx_bias; 3883 extern int dev_weight_tx_bias; 3884 extern int dev_rx_weight; 3885 extern int dev_tx_weight; 3886 3887 bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev); 3888 struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev, 3889 struct list_head **iter); 3890 struct net_device *netdev_all_upper_get_next_dev_rcu(struct net_device *dev, 3891 struct list_head **iter); 3892 3893 /* iterate through upper list, must be called under RCU read lock */ 3894 #define netdev_for_each_upper_dev_rcu(dev, updev, iter) \ 3895 for (iter = &(dev)->adj_list.upper, \ 3896 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \ 3897 updev; \ 3898 updev = netdev_upper_get_next_dev_rcu(dev, &(iter))) 3899 3900 int netdev_walk_all_upper_dev_rcu(struct net_device *dev, 3901 int (*fn)(struct net_device *upper_dev, 3902 void *data), 3903 void *data); 3904 3905 bool netdev_has_upper_dev_all_rcu(struct net_device *dev, 3906 struct net_device *upper_dev); 3907 3908 bool netdev_has_any_upper_dev(struct net_device *dev); 3909 3910 void *netdev_lower_get_next_private(struct net_device *dev, 3911 struct list_head **iter); 3912 void *netdev_lower_get_next_private_rcu(struct net_device *dev, 3913 struct list_head **iter); 3914 3915 #define netdev_for_each_lower_private(dev, priv, iter) \ 3916 for (iter = (dev)->adj_list.lower.next, \ 3917 priv = netdev_lower_get_next_private(dev, &(iter)); \ 3918 priv; \ 3919 priv = netdev_lower_get_next_private(dev, &(iter))) 3920 3921 #define netdev_for_each_lower_private_rcu(dev, priv, iter) \ 3922 for (iter = &(dev)->adj_list.lower, \ 3923 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \ 3924 priv; \ 3925 priv = netdev_lower_get_next_private_rcu(dev, &(iter))) 3926 3927 void *netdev_lower_get_next(struct net_device *dev, 3928 struct list_head **iter); 3929 3930 #define netdev_for_each_lower_dev(dev, ldev, iter) \ 3931 for (iter = (dev)->adj_list.lower.next, \ 3932 ldev = netdev_lower_get_next(dev, &(iter)); \ 3933 ldev; \ 3934 ldev = netdev_lower_get_next(dev, &(iter))) 3935 3936 struct net_device *netdev_all_lower_get_next(struct net_device *dev, 3937 struct list_head **iter); 3938 struct net_device *netdev_all_lower_get_next_rcu(struct net_device *dev, 3939 struct list_head **iter); 3940 3941 int netdev_walk_all_lower_dev(struct net_device *dev, 3942 int (*fn)(struct net_device *lower_dev, 3943 void *data), 3944 void *data); 3945 int netdev_walk_all_lower_dev_rcu(struct net_device *dev, 3946 int (*fn)(struct net_device *lower_dev, 3947 void *data), 3948 void *data); 3949 3950 void *netdev_adjacent_get_private(struct list_head *adj_list); 3951 void *netdev_lower_get_first_private_rcu(struct net_device *dev); 3952 struct net_device *netdev_master_upper_dev_get(struct net_device *dev); 3953 struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev); 3954 int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev, 3955 struct netlink_ext_ack *extack); 3956 int netdev_master_upper_dev_link(struct net_device *dev, 3957 struct net_device *upper_dev, 3958 void *upper_priv, void *upper_info, 3959 struct netlink_ext_ack *extack); 3960 void netdev_upper_dev_unlink(struct net_device *dev, 3961 struct net_device *upper_dev); 3962 void netdev_adjacent_rename_links(struct net_device *dev, char *oldname); 3963 void *netdev_lower_dev_get_private(struct net_device *dev, 3964 struct net_device *lower_dev); 3965 void netdev_lower_state_changed(struct net_device *lower_dev, 3966 void *lower_state_info); 3967 3968 /* RSS keys are 40 or 52 bytes long */ 3969 #define NETDEV_RSS_KEY_LEN 52 3970 extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN] __read_mostly; 3971 void netdev_rss_key_fill(void *buffer, size_t len); 3972 3973 int dev_get_nest_level(struct net_device *dev); 3974 int skb_checksum_help(struct sk_buff *skb); 3975 int skb_crc32c_csum_help(struct sk_buff *skb); 3976 int skb_csum_hwoffload_help(struct sk_buff *skb, 3977 const netdev_features_t features); 3978 3979 struct sk_buff *__skb_gso_segment(struct sk_buff *skb, 3980 netdev_features_t features, bool tx_path); 3981 struct sk_buff *skb_mac_gso_segment(struct sk_buff *skb, 3982 netdev_features_t features); 3983 3984 struct netdev_bonding_info { 3985 ifslave slave; 3986 ifbond master; 3987 }; 3988 3989 struct netdev_notifier_bonding_info { 3990 struct netdev_notifier_info info; /* must be first */ 3991 struct netdev_bonding_info bonding_info; 3992 }; 3993 3994 void netdev_bonding_info_change(struct net_device *dev, 3995 struct netdev_bonding_info *bonding_info); 3996 3997 static inline 3998 struct sk_buff *skb_gso_segment(struct sk_buff *skb, netdev_features_t features) 3999 { 4000 return __skb_gso_segment(skb, features, true); 4001 } 4002 __be16 skb_network_protocol(struct sk_buff *skb, int *depth); 4003 4004 static inline bool can_checksum_protocol(netdev_features_t features, 4005 __be16 protocol) 4006 { 4007 if (protocol == htons(ETH_P_FCOE)) 4008 return !!(features & NETIF_F_FCOE_CRC); 4009 4010 /* Assume this is an IP checksum (not SCTP CRC) */ 4011 4012 if (features & NETIF_F_HW_CSUM) { 4013 /* Can checksum everything */ 4014 return true; 4015 } 4016 4017 switch (protocol) { 4018 case htons(ETH_P_IP): 4019 return !!(features & NETIF_F_IP_CSUM); 4020 case htons(ETH_P_IPV6): 4021 return !!(features & NETIF_F_IPV6_CSUM); 4022 default: 4023 return false; 4024 } 4025 } 4026 4027 #ifdef CONFIG_BUG 4028 void netdev_rx_csum_fault(struct net_device *dev); 4029 #else 4030 static inline void netdev_rx_csum_fault(struct net_device *dev) 4031 { 4032 } 4033 #endif 4034 /* rx skb timestamps */ 4035 void net_enable_timestamp(void); 4036 void net_disable_timestamp(void); 4037 4038 #ifdef CONFIG_PROC_FS 4039 int __init dev_proc_init(void); 4040 #else 4041 #define dev_proc_init() 0 4042 #endif 4043 4044 static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops, 4045 struct sk_buff *skb, struct net_device *dev, 4046 bool more) 4047 { 4048 skb->xmit_more = more ? 1 : 0; 4049 return ops->ndo_start_xmit(skb, dev); 4050 } 4051 4052 static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev, 4053 struct netdev_queue *txq, bool more) 4054 { 4055 const struct net_device_ops *ops = dev->netdev_ops; 4056 int rc; 4057 4058 rc = __netdev_start_xmit(ops, skb, dev, more); 4059 if (rc == NETDEV_TX_OK) 4060 txq_trans_update(txq); 4061 4062 return rc; 4063 } 4064 4065 int netdev_class_create_file_ns(const struct class_attribute *class_attr, 4066 const void *ns); 4067 void netdev_class_remove_file_ns(const struct class_attribute *class_attr, 4068 const void *ns); 4069 4070 static inline int netdev_class_create_file(const struct class_attribute *class_attr) 4071 { 4072 return netdev_class_create_file_ns(class_attr, NULL); 4073 } 4074 4075 static inline void netdev_class_remove_file(const struct class_attribute *class_attr) 4076 { 4077 netdev_class_remove_file_ns(class_attr, NULL); 4078 } 4079 4080 extern const struct kobj_ns_type_operations net_ns_type_operations; 4081 4082 const char *netdev_drivername(const struct net_device *dev); 4083 4084 void linkwatch_run_queue(void); 4085 4086 static inline netdev_features_t netdev_intersect_features(netdev_features_t f1, 4087 netdev_features_t f2) 4088 { 4089 if ((f1 ^ f2) & NETIF_F_HW_CSUM) { 4090 if (f1 & NETIF_F_HW_CSUM) 4091 f1 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM); 4092 else 4093 f2 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM); 4094 } 4095 4096 return f1 & f2; 4097 } 4098 4099 static inline netdev_features_t netdev_get_wanted_features( 4100 struct net_device *dev) 4101 { 4102 return (dev->features & ~dev->hw_features) | dev->wanted_features; 4103 } 4104 netdev_features_t netdev_increment_features(netdev_features_t all, 4105 netdev_features_t one, netdev_features_t mask); 4106 4107 /* Allow TSO being used on stacked device : 4108 * Performing the GSO segmentation before last device 4109 * is a performance improvement. 4110 */ 4111 static inline netdev_features_t netdev_add_tso_features(netdev_features_t features, 4112 netdev_features_t mask) 4113 { 4114 return netdev_increment_features(features, NETIF_F_ALL_TSO, mask); 4115 } 4116 4117 int __netdev_update_features(struct net_device *dev); 4118 void netdev_update_features(struct net_device *dev); 4119 void netdev_change_features(struct net_device *dev); 4120 4121 void netif_stacked_transfer_operstate(const struct net_device *rootdev, 4122 struct net_device *dev); 4123 4124 netdev_features_t passthru_features_check(struct sk_buff *skb, 4125 struct net_device *dev, 4126 netdev_features_t features); 4127 netdev_features_t netif_skb_features(struct sk_buff *skb); 4128 4129 static inline bool net_gso_ok(netdev_features_t features, int gso_type) 4130 { 4131 netdev_features_t feature = (netdev_features_t)gso_type << NETIF_F_GSO_SHIFT; 4132 4133 /* check flags correspondence */ 4134 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT)); 4135 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT)); 4136 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT)); 4137 BUILD_BUG_ON(SKB_GSO_TCP_FIXEDID != (NETIF_F_TSO_MANGLEID >> NETIF_F_GSO_SHIFT)); 4138 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT)); 4139 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT)); 4140 BUILD_BUG_ON(SKB_GSO_GRE != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT)); 4141 BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT)); 4142 BUILD_BUG_ON(SKB_GSO_IPXIP4 != (NETIF_F_GSO_IPXIP4 >> NETIF_F_GSO_SHIFT)); 4143 BUILD_BUG_ON(SKB_GSO_IPXIP6 != (NETIF_F_GSO_IPXIP6 >> NETIF_F_GSO_SHIFT)); 4144 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT)); 4145 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT)); 4146 BUILD_BUG_ON(SKB_GSO_PARTIAL != (NETIF_F_GSO_PARTIAL >> NETIF_F_GSO_SHIFT)); 4147 BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT)); 4148 BUILD_BUG_ON(SKB_GSO_SCTP != (NETIF_F_GSO_SCTP >> NETIF_F_GSO_SHIFT)); 4149 BUILD_BUG_ON(SKB_GSO_ESP != (NETIF_F_GSO_ESP >> NETIF_F_GSO_SHIFT)); 4150 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_GSO_UDP >> NETIF_F_GSO_SHIFT)); 4151 4152 return (features & feature) == feature; 4153 } 4154 4155 static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features) 4156 { 4157 return net_gso_ok(features, skb_shinfo(skb)->gso_type) && 4158 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST)); 4159 } 4160 4161 static inline bool netif_needs_gso(struct sk_buff *skb, 4162 netdev_features_t features) 4163 { 4164 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) || 4165 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) && 4166 (skb->ip_summed != CHECKSUM_UNNECESSARY))); 4167 } 4168 4169 static inline void netif_set_gso_max_size(struct net_device *dev, 4170 unsigned int size) 4171 { 4172 dev->gso_max_size = size; 4173 } 4174 4175 static inline void skb_gso_error_unwind(struct sk_buff *skb, __be16 protocol, 4176 int pulled_hlen, u16 mac_offset, 4177 int mac_len) 4178 { 4179 skb->protocol = protocol; 4180 skb->encapsulation = 1; 4181 skb_push(skb, pulled_hlen); 4182 skb_reset_transport_header(skb); 4183 skb->mac_header = mac_offset; 4184 skb->network_header = skb->mac_header + mac_len; 4185 skb->mac_len = mac_len; 4186 } 4187 4188 static inline bool netif_is_macsec(const struct net_device *dev) 4189 { 4190 return dev->priv_flags & IFF_MACSEC; 4191 } 4192 4193 static inline bool netif_is_macvlan(const struct net_device *dev) 4194 { 4195 return dev->priv_flags & IFF_MACVLAN; 4196 } 4197 4198 static inline bool netif_is_macvlan_port(const struct net_device *dev) 4199 { 4200 return dev->priv_flags & IFF_MACVLAN_PORT; 4201 } 4202 4203 static inline bool netif_is_ipvlan(const struct net_device *dev) 4204 { 4205 return dev->priv_flags & IFF_IPVLAN_SLAVE; 4206 } 4207 4208 static inline bool netif_is_ipvlan_port(const struct net_device *dev) 4209 { 4210 return dev->priv_flags & IFF_IPVLAN_MASTER; 4211 } 4212 4213 static inline bool netif_is_bond_master(const struct net_device *dev) 4214 { 4215 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING; 4216 } 4217 4218 static inline bool netif_is_bond_slave(const struct net_device *dev) 4219 { 4220 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING; 4221 } 4222 4223 static inline bool netif_supports_nofcs(struct net_device *dev) 4224 { 4225 return dev->priv_flags & IFF_SUPP_NOFCS; 4226 } 4227 4228 static inline bool netif_is_l3_master(const struct net_device *dev) 4229 { 4230 return dev->priv_flags & IFF_L3MDEV_MASTER; 4231 } 4232 4233 static inline bool netif_is_l3_slave(const struct net_device *dev) 4234 { 4235 return dev->priv_flags & IFF_L3MDEV_SLAVE; 4236 } 4237 4238 static inline bool netif_is_bridge_master(const struct net_device *dev) 4239 { 4240 return dev->priv_flags & IFF_EBRIDGE; 4241 } 4242 4243 static inline bool netif_is_bridge_port(const struct net_device *dev) 4244 { 4245 return dev->priv_flags & IFF_BRIDGE_PORT; 4246 } 4247 4248 static inline bool netif_is_ovs_master(const struct net_device *dev) 4249 { 4250 return dev->priv_flags & IFF_OPENVSWITCH; 4251 } 4252 4253 static inline bool netif_is_ovs_port(const struct net_device *dev) 4254 { 4255 return dev->priv_flags & IFF_OVS_DATAPATH; 4256 } 4257 4258 static inline bool netif_is_team_master(const struct net_device *dev) 4259 { 4260 return dev->priv_flags & IFF_TEAM; 4261 } 4262 4263 static inline bool netif_is_team_port(const struct net_device *dev) 4264 { 4265 return dev->priv_flags & IFF_TEAM_PORT; 4266 } 4267 4268 static inline bool netif_is_lag_master(const struct net_device *dev) 4269 { 4270 return netif_is_bond_master(dev) || netif_is_team_master(dev); 4271 } 4272 4273 static inline bool netif_is_lag_port(const struct net_device *dev) 4274 { 4275 return netif_is_bond_slave(dev) || netif_is_team_port(dev); 4276 } 4277 4278 static inline bool netif_is_rxfh_configured(const struct net_device *dev) 4279 { 4280 return dev->priv_flags & IFF_RXFH_CONFIGURED; 4281 } 4282 4283 /* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */ 4284 static inline void netif_keep_dst(struct net_device *dev) 4285 { 4286 dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM); 4287 } 4288 4289 /* return true if dev can't cope with mtu frames that need vlan tag insertion */ 4290 static inline bool netif_reduces_vlan_mtu(struct net_device *dev) 4291 { 4292 /* TODO: reserve and use an additional IFF bit, if we get more users */ 4293 return dev->priv_flags & IFF_MACSEC; 4294 } 4295 4296 extern struct pernet_operations __net_initdata loopback_net_ops; 4297 4298 /* Logging, debugging and troubleshooting/diagnostic helpers. */ 4299 4300 /* netdev_printk helpers, similar to dev_printk */ 4301 4302 static inline const char *netdev_name(const struct net_device *dev) 4303 { 4304 if (!dev->name[0] || strchr(dev->name, '%')) 4305 return "(unnamed net_device)"; 4306 return dev->name; 4307 } 4308 4309 static inline bool netdev_unregistering(const struct net_device *dev) 4310 { 4311 return dev->reg_state == NETREG_UNREGISTERING; 4312 } 4313 4314 static inline const char *netdev_reg_state(const struct net_device *dev) 4315 { 4316 switch (dev->reg_state) { 4317 case NETREG_UNINITIALIZED: return " (uninitialized)"; 4318 case NETREG_REGISTERED: return ""; 4319 case NETREG_UNREGISTERING: return " (unregistering)"; 4320 case NETREG_UNREGISTERED: return " (unregistered)"; 4321 case NETREG_RELEASED: return " (released)"; 4322 case NETREG_DUMMY: return " (dummy)"; 4323 } 4324 4325 WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, dev->reg_state); 4326 return " (unknown)"; 4327 } 4328 4329 __printf(3, 4) 4330 void netdev_printk(const char *level, const struct net_device *dev, 4331 const char *format, ...); 4332 __printf(2, 3) 4333 void netdev_emerg(const struct net_device *dev, const char *format, ...); 4334 __printf(2, 3) 4335 void netdev_alert(const struct net_device *dev, const char *format, ...); 4336 __printf(2, 3) 4337 void netdev_crit(const struct net_device *dev, const char *format, ...); 4338 __printf(2, 3) 4339 void netdev_err(const struct net_device *dev, const char *format, ...); 4340 __printf(2, 3) 4341 void netdev_warn(const struct net_device *dev, const char *format, ...); 4342 __printf(2, 3) 4343 void netdev_notice(const struct net_device *dev, const char *format, ...); 4344 __printf(2, 3) 4345 void netdev_info(const struct net_device *dev, const char *format, ...); 4346 4347 #define netdev_level_once(level, dev, fmt, ...) \ 4348 do { \ 4349 static bool __print_once __read_mostly; \ 4350 \ 4351 if (!__print_once) { \ 4352 __print_once = true; \ 4353 netdev_printk(level, dev, fmt, ##__VA_ARGS__); \ 4354 } \ 4355 } while (0) 4356 4357 #define netdev_emerg_once(dev, fmt, ...) \ 4358 netdev_level_once(KERN_EMERG, dev, fmt, ##__VA_ARGS__) 4359 #define netdev_alert_once(dev, fmt, ...) \ 4360 netdev_level_once(KERN_ALERT, dev, fmt, ##__VA_ARGS__) 4361 #define netdev_crit_once(dev, fmt, ...) \ 4362 netdev_level_once(KERN_CRIT, dev, fmt, ##__VA_ARGS__) 4363 #define netdev_err_once(dev, fmt, ...) \ 4364 netdev_level_once(KERN_ERR, dev, fmt, ##__VA_ARGS__) 4365 #define netdev_warn_once(dev, fmt, ...) \ 4366 netdev_level_once(KERN_WARNING, dev, fmt, ##__VA_ARGS__) 4367 #define netdev_notice_once(dev, fmt, ...) \ 4368 netdev_level_once(KERN_NOTICE, dev, fmt, ##__VA_ARGS__) 4369 #define netdev_info_once(dev, fmt, ...) \ 4370 netdev_level_once(KERN_INFO, dev, fmt, ##__VA_ARGS__) 4371 4372 #define MODULE_ALIAS_NETDEV(device) \ 4373 MODULE_ALIAS("netdev-" device) 4374 4375 #if defined(CONFIG_DYNAMIC_DEBUG) 4376 #define netdev_dbg(__dev, format, args...) \ 4377 do { \ 4378 dynamic_netdev_dbg(__dev, format, ##args); \ 4379 } while (0) 4380 #elif defined(DEBUG) 4381 #define netdev_dbg(__dev, format, args...) \ 4382 netdev_printk(KERN_DEBUG, __dev, format, ##args) 4383 #else 4384 #define netdev_dbg(__dev, format, args...) \ 4385 ({ \ 4386 if (0) \ 4387 netdev_printk(KERN_DEBUG, __dev, format, ##args); \ 4388 }) 4389 #endif 4390 4391 #if defined(VERBOSE_DEBUG) 4392 #define netdev_vdbg netdev_dbg 4393 #else 4394 4395 #define netdev_vdbg(dev, format, args...) \ 4396 ({ \ 4397 if (0) \ 4398 netdev_printk(KERN_DEBUG, dev, format, ##args); \ 4399 0; \ 4400 }) 4401 #endif 4402 4403 /* 4404 * netdev_WARN() acts like dev_printk(), but with the key difference 4405 * of using a WARN/WARN_ON to get the message out, including the 4406 * file/line information and a backtrace. 4407 */ 4408 #define netdev_WARN(dev, format, args...) \ 4409 WARN(1, "netdevice: %s%s\n" format, netdev_name(dev), \ 4410 netdev_reg_state(dev), ##args) 4411 4412 #define netdev_WARN_ONCE(dev, condition, format, arg...) \ 4413 WARN_ONCE(1, "netdevice: %s%s\n" format, netdev_name(dev) \ 4414 netdev_reg_state(dev), ##args) 4415 4416 /* netif printk helpers, similar to netdev_printk */ 4417 4418 #define netif_printk(priv, type, level, dev, fmt, args...) \ 4419 do { \ 4420 if (netif_msg_##type(priv)) \ 4421 netdev_printk(level, (dev), fmt, ##args); \ 4422 } while (0) 4423 4424 #define netif_level(level, priv, type, dev, fmt, args...) \ 4425 do { \ 4426 if (netif_msg_##type(priv)) \ 4427 netdev_##level(dev, fmt, ##args); \ 4428 } while (0) 4429 4430 #define netif_emerg(priv, type, dev, fmt, args...) \ 4431 netif_level(emerg, priv, type, dev, fmt, ##args) 4432 #define netif_alert(priv, type, dev, fmt, args...) \ 4433 netif_level(alert, priv, type, dev, fmt, ##args) 4434 #define netif_crit(priv, type, dev, fmt, args...) \ 4435 netif_level(crit, priv, type, dev, fmt, ##args) 4436 #define netif_err(priv, type, dev, fmt, args...) \ 4437 netif_level(err, priv, type, dev, fmt, ##args) 4438 #define netif_warn(priv, type, dev, fmt, args...) \ 4439 netif_level(warn, priv, type, dev, fmt, ##args) 4440 #define netif_notice(priv, type, dev, fmt, args...) \ 4441 netif_level(notice, priv, type, dev, fmt, ##args) 4442 #define netif_info(priv, type, dev, fmt, args...) \ 4443 netif_level(info, priv, type, dev, fmt, ##args) 4444 4445 #if defined(CONFIG_DYNAMIC_DEBUG) 4446 #define netif_dbg(priv, type, netdev, format, args...) \ 4447 do { \ 4448 if (netif_msg_##type(priv)) \ 4449 dynamic_netdev_dbg(netdev, format, ##args); \ 4450 } while (0) 4451 #elif defined(DEBUG) 4452 #define netif_dbg(priv, type, dev, format, args...) \ 4453 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args) 4454 #else 4455 #define netif_dbg(priv, type, dev, format, args...) \ 4456 ({ \ 4457 if (0) \ 4458 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \ 4459 0; \ 4460 }) 4461 #endif 4462 4463 /* if @cond then downgrade to debug, else print at @level */ 4464 #define netif_cond_dbg(priv, type, netdev, cond, level, fmt, args...) \ 4465 do { \ 4466 if (cond) \ 4467 netif_dbg(priv, type, netdev, fmt, ##args); \ 4468 else \ 4469 netif_ ## level(priv, type, netdev, fmt, ##args); \ 4470 } while (0) 4471 4472 #if defined(VERBOSE_DEBUG) 4473 #define netif_vdbg netif_dbg 4474 #else 4475 #define netif_vdbg(priv, type, dev, format, args...) \ 4476 ({ \ 4477 if (0) \ 4478 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \ 4479 0; \ 4480 }) 4481 #endif 4482 4483 /* 4484 * The list of packet types we will receive (as opposed to discard) 4485 * and the routines to invoke. 4486 * 4487 * Why 16. Because with 16 the only overlap we get on a hash of the 4488 * low nibble of the protocol value is RARP/SNAP/X.25. 4489 * 4490 * 0800 IP 4491 * 0001 802.3 4492 * 0002 AX.25 4493 * 0004 802.2 4494 * 8035 RARP 4495 * 0005 SNAP 4496 * 0805 X.25 4497 * 0806 ARP 4498 * 8137 IPX 4499 * 0009 Localtalk 4500 * 86DD IPv6 4501 */ 4502 #define PTYPE_HASH_SIZE (16) 4503 #define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1) 4504 4505 #endif /* _LINUX_NETDEVICE_H */ 4506