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 48 #include <linux/netdev_features.h> 49 #include <linux/neighbour.h> 50 #include <uapi/linux/netdevice.h> 51 #include <uapi/linux/if_bonding.h> 52 #include <uapi/linux/pkt_cls.h> 53 #include <linux/hashtable.h> 54 55 struct netpoll_info; 56 struct device; 57 struct phy_device; 58 struct dsa_port; 59 60 /* 802.11 specific */ 61 struct wireless_dev; 62 /* 802.15.4 specific */ 63 struct wpan_dev; 64 struct mpls_dev; 65 /* UDP Tunnel offloads */ 66 struct udp_tunnel_info; 67 struct bpf_prog; 68 struct xdp_buff; 69 70 void netdev_set_default_ethtool_ops(struct net_device *dev, 71 const struct ethtool_ops *ops); 72 73 /* Backlog congestion levels */ 74 #define NET_RX_SUCCESS 0 /* keep 'em coming, baby */ 75 #define NET_RX_DROP 1 /* packet dropped */ 76 77 /* 78 * Transmit return codes: transmit return codes originate from three different 79 * namespaces: 80 * 81 * - qdisc return codes 82 * - driver transmit return codes 83 * - errno values 84 * 85 * Drivers are allowed to return any one of those in their hard_start_xmit() 86 * function. Real network devices commonly used with qdiscs should only return 87 * the driver transmit return codes though - when qdiscs are used, the actual 88 * transmission happens asynchronously, so the value is not propagated to 89 * higher layers. Virtual network devices transmit synchronously; in this case 90 * the driver transmit return codes are consumed by dev_queue_xmit(), and all 91 * others are propagated to higher layers. 92 */ 93 94 /* qdisc ->enqueue() return codes. */ 95 #define NET_XMIT_SUCCESS 0x00 96 #define NET_XMIT_DROP 0x01 /* skb dropped */ 97 #define NET_XMIT_CN 0x02 /* congestion notification */ 98 #define NET_XMIT_MASK 0x0f /* qdisc flags in net/sch_generic.h */ 99 100 /* NET_XMIT_CN is special. It does not guarantee that this packet is lost. It 101 * indicates that the device will soon be dropping packets, or already drops 102 * some packets of the same priority; prompting us to send less aggressively. */ 103 #define net_xmit_eval(e) ((e) == NET_XMIT_CN ? 0 : (e)) 104 #define net_xmit_errno(e) ((e) != NET_XMIT_CN ? -ENOBUFS : 0) 105 106 /* Driver transmit return codes */ 107 #define NETDEV_TX_MASK 0xf0 108 109 enum netdev_tx { 110 __NETDEV_TX_MIN = INT_MIN, /* make sure enum is signed */ 111 NETDEV_TX_OK = 0x00, /* driver took care of packet */ 112 NETDEV_TX_BUSY = 0x10, /* driver tx path was busy*/ 113 }; 114 typedef enum netdev_tx netdev_tx_t; 115 116 /* 117 * Current order: NETDEV_TX_MASK > NET_XMIT_MASK >= 0 is significant; 118 * hard_start_xmit() return < NET_XMIT_MASK means skb was consumed. 119 */ 120 static inline bool dev_xmit_complete(int rc) 121 { 122 /* 123 * Positive cases with an skb consumed by a driver: 124 * - successful transmission (rc == NETDEV_TX_OK) 125 * - error while transmitting (rc < 0) 126 * - error while queueing to a different device (rc & NET_XMIT_MASK) 127 */ 128 if (likely(rc < NET_XMIT_MASK)) 129 return true; 130 131 return false; 132 } 133 134 /* 135 * Compute the worst-case header length according to the protocols 136 * used. 137 */ 138 139 #if defined(CONFIG_HYPERV_NET) 140 # define LL_MAX_HEADER 128 141 #elif defined(CONFIG_WLAN) || IS_ENABLED(CONFIG_AX25) 142 # if defined(CONFIG_MAC80211_MESH) 143 # define LL_MAX_HEADER 128 144 # else 145 # define LL_MAX_HEADER 96 146 # endif 147 #else 148 # define LL_MAX_HEADER 32 149 #endif 150 151 #if !IS_ENABLED(CONFIG_NET_IPIP) && !IS_ENABLED(CONFIG_NET_IPGRE) && \ 152 !IS_ENABLED(CONFIG_IPV6_SIT) && !IS_ENABLED(CONFIG_IPV6_TUNNEL) 153 #define MAX_HEADER LL_MAX_HEADER 154 #else 155 #define MAX_HEADER (LL_MAX_HEADER + 48) 156 #endif 157 158 /* 159 * Old network device statistics. Fields are native words 160 * (unsigned long) so they can be read and written atomically. 161 */ 162 163 struct net_device_stats { 164 unsigned long rx_packets; 165 unsigned long tx_packets; 166 unsigned long rx_bytes; 167 unsigned long tx_bytes; 168 unsigned long rx_errors; 169 unsigned long tx_errors; 170 unsigned long rx_dropped; 171 unsigned long tx_dropped; 172 unsigned long multicast; 173 unsigned long collisions; 174 unsigned long rx_length_errors; 175 unsigned long rx_over_errors; 176 unsigned long rx_crc_errors; 177 unsigned long rx_frame_errors; 178 unsigned long rx_fifo_errors; 179 unsigned long rx_missed_errors; 180 unsigned long tx_aborted_errors; 181 unsigned long tx_carrier_errors; 182 unsigned long tx_fifo_errors; 183 unsigned long tx_heartbeat_errors; 184 unsigned long tx_window_errors; 185 unsigned long rx_compressed; 186 unsigned long tx_compressed; 187 }; 188 189 190 #include <linux/cache.h> 191 #include <linux/skbuff.h> 192 193 #ifdef CONFIG_RPS 194 #include <linux/static_key.h> 195 extern struct static_key rps_needed; 196 extern struct static_key rfs_needed; 197 #endif 198 199 struct neighbour; 200 struct neigh_parms; 201 struct sk_buff; 202 203 struct netdev_hw_addr { 204 struct list_head list; 205 unsigned char addr[MAX_ADDR_LEN]; 206 unsigned char type; 207 #define NETDEV_HW_ADDR_T_LAN 1 208 #define NETDEV_HW_ADDR_T_SAN 2 209 #define NETDEV_HW_ADDR_T_SLAVE 3 210 #define NETDEV_HW_ADDR_T_UNICAST 4 211 #define NETDEV_HW_ADDR_T_MULTICAST 5 212 bool global_use; 213 int sync_cnt; 214 int refcount; 215 int synced; 216 struct rcu_head rcu_head; 217 }; 218 219 struct netdev_hw_addr_list { 220 struct list_head list; 221 int count; 222 }; 223 224 #define netdev_hw_addr_list_count(l) ((l)->count) 225 #define netdev_hw_addr_list_empty(l) (netdev_hw_addr_list_count(l) == 0) 226 #define netdev_hw_addr_list_for_each(ha, l) \ 227 list_for_each_entry(ha, &(l)->list, list) 228 229 #define netdev_uc_count(dev) netdev_hw_addr_list_count(&(dev)->uc) 230 #define netdev_uc_empty(dev) netdev_hw_addr_list_empty(&(dev)->uc) 231 #define netdev_for_each_uc_addr(ha, dev) \ 232 netdev_hw_addr_list_for_each(ha, &(dev)->uc) 233 234 #define netdev_mc_count(dev) netdev_hw_addr_list_count(&(dev)->mc) 235 #define netdev_mc_empty(dev) netdev_hw_addr_list_empty(&(dev)->mc) 236 #define netdev_for_each_mc_addr(ha, dev) \ 237 netdev_hw_addr_list_for_each(ha, &(dev)->mc) 238 239 struct hh_cache { 240 unsigned int hh_len; 241 seqlock_t hh_lock; 242 243 /* cached hardware header; allow for machine alignment needs. */ 244 #define HH_DATA_MOD 16 245 #define HH_DATA_OFF(__len) \ 246 (HH_DATA_MOD - (((__len - 1) & (HH_DATA_MOD - 1)) + 1)) 247 #define HH_DATA_ALIGN(__len) \ 248 (((__len)+(HH_DATA_MOD-1))&~(HH_DATA_MOD - 1)) 249 unsigned long hh_data[HH_DATA_ALIGN(LL_MAX_HEADER) / sizeof(long)]; 250 }; 251 252 /* Reserve HH_DATA_MOD byte-aligned hard_header_len, but at least that much. 253 * Alternative is: 254 * dev->hard_header_len ? (dev->hard_header_len + 255 * (HH_DATA_MOD - 1)) & ~(HH_DATA_MOD - 1) : 0 256 * 257 * We could use other alignment values, but we must maintain the 258 * relationship HH alignment <= LL alignment. 259 */ 260 #define LL_RESERVED_SPACE(dev) \ 261 ((((dev)->hard_header_len+(dev)->needed_headroom)&~(HH_DATA_MOD - 1)) + HH_DATA_MOD) 262 #define LL_RESERVED_SPACE_EXTRA(dev,extra) \ 263 ((((dev)->hard_header_len+(dev)->needed_headroom+(extra))&~(HH_DATA_MOD - 1)) + HH_DATA_MOD) 264 265 struct header_ops { 266 int (*create) (struct sk_buff *skb, struct net_device *dev, 267 unsigned short type, const void *daddr, 268 const void *saddr, unsigned int len); 269 int (*parse)(const struct sk_buff *skb, unsigned char *haddr); 270 int (*cache)(const struct neighbour *neigh, struct hh_cache *hh, __be16 type); 271 void (*cache_update)(struct hh_cache *hh, 272 const struct net_device *dev, 273 const unsigned char *haddr); 274 bool (*validate)(const char *ll_header, unsigned int len); 275 }; 276 277 /* These flag bits are private to the generic network queueing 278 * layer; they may not be explicitly referenced by any other 279 * code. 280 */ 281 282 enum netdev_state_t { 283 __LINK_STATE_START, 284 __LINK_STATE_PRESENT, 285 __LINK_STATE_NOCARRIER, 286 __LINK_STATE_LINKWATCH_PENDING, 287 __LINK_STATE_DORMANT, 288 }; 289 290 291 /* 292 * This structure holds boot-time configured netdevice settings. They 293 * are then used in the device probing. 294 */ 295 struct netdev_boot_setup { 296 char name[IFNAMSIZ]; 297 struct ifmap map; 298 }; 299 #define NETDEV_BOOT_SETUP_MAX 8 300 301 int __init netdev_boot_setup(char *str); 302 303 /* 304 * Structure for NAPI scheduling similar to tasklet but with weighting 305 */ 306 struct napi_struct { 307 /* The poll_list must only be managed by the entity which 308 * changes the state of the NAPI_STATE_SCHED bit. This means 309 * whoever atomically sets that bit can add this napi_struct 310 * to the per-CPU poll_list, and whoever clears that bit 311 * can remove from the list right before clearing the bit. 312 */ 313 struct list_head poll_list; 314 315 unsigned long state; 316 int weight; 317 unsigned int gro_count; 318 int (*poll)(struct napi_struct *, int); 319 #ifdef CONFIG_NETPOLL 320 int poll_owner; 321 #endif 322 struct net_device *dev; 323 struct sk_buff *gro_list; 324 struct sk_buff *skb; 325 struct hrtimer timer; 326 struct list_head dev_list; 327 struct hlist_node napi_hash_node; 328 unsigned int napi_id; 329 }; 330 331 enum { 332 NAPI_STATE_SCHED, /* Poll is scheduled */ 333 NAPI_STATE_MISSED, /* reschedule a napi */ 334 NAPI_STATE_DISABLE, /* Disable pending */ 335 NAPI_STATE_NPSVC, /* Netpoll - don't dequeue from poll_list */ 336 NAPI_STATE_HASHED, /* In NAPI hash (busy polling possible) */ 337 NAPI_STATE_NO_BUSY_POLL,/* Do not add in napi_hash, no busy polling */ 338 NAPI_STATE_IN_BUSY_POLL,/* sk_busy_loop() owns this NAPI */ 339 }; 340 341 enum { 342 NAPIF_STATE_SCHED = BIT(NAPI_STATE_SCHED), 343 NAPIF_STATE_MISSED = BIT(NAPI_STATE_MISSED), 344 NAPIF_STATE_DISABLE = BIT(NAPI_STATE_DISABLE), 345 NAPIF_STATE_NPSVC = BIT(NAPI_STATE_NPSVC), 346 NAPIF_STATE_HASHED = BIT(NAPI_STATE_HASHED), 347 NAPIF_STATE_NO_BUSY_POLL = BIT(NAPI_STATE_NO_BUSY_POLL), 348 NAPIF_STATE_IN_BUSY_POLL = BIT(NAPI_STATE_IN_BUSY_POLL), 349 }; 350 351 enum gro_result { 352 GRO_MERGED, 353 GRO_MERGED_FREE, 354 GRO_HELD, 355 GRO_NORMAL, 356 GRO_DROP, 357 GRO_CONSUMED, 358 }; 359 typedef enum gro_result gro_result_t; 360 361 /* 362 * enum rx_handler_result - Possible return values for rx_handlers. 363 * @RX_HANDLER_CONSUMED: skb was consumed by rx_handler, do not process it 364 * further. 365 * @RX_HANDLER_ANOTHER: Do another round in receive path. This is indicated in 366 * case skb->dev was changed by rx_handler. 367 * @RX_HANDLER_EXACT: Force exact delivery, no wildcard. 368 * @RX_HANDLER_PASS: Do nothing, pass the skb as if no rx_handler was called. 369 * 370 * rx_handlers are functions called from inside __netif_receive_skb(), to do 371 * special processing of the skb, prior to delivery to protocol handlers. 372 * 373 * Currently, a net_device can only have a single rx_handler registered. Trying 374 * to register a second rx_handler will return -EBUSY. 375 * 376 * To register a rx_handler on a net_device, use netdev_rx_handler_register(). 377 * To unregister a rx_handler on a net_device, use 378 * netdev_rx_handler_unregister(). 379 * 380 * Upon return, rx_handler is expected to tell __netif_receive_skb() what to 381 * do with the skb. 382 * 383 * If the rx_handler consumed the skb in some way, it should return 384 * RX_HANDLER_CONSUMED. This is appropriate when the rx_handler arranged for 385 * the skb to be delivered in some other way. 386 * 387 * If the rx_handler changed skb->dev, to divert the skb to another 388 * net_device, it should return RX_HANDLER_ANOTHER. The rx_handler for the 389 * new device will be called if it exists. 390 * 391 * If the rx_handler decides the skb should be ignored, it should return 392 * RX_HANDLER_EXACT. The skb will only be delivered to protocol handlers that 393 * are registered on exact device (ptype->dev == skb->dev). 394 * 395 * If the rx_handler didn't change skb->dev, but wants the skb to be normally 396 * delivered, it should return RX_HANDLER_PASS. 397 * 398 * A device without a registered rx_handler will behave as if rx_handler 399 * returned RX_HANDLER_PASS. 400 */ 401 402 enum rx_handler_result { 403 RX_HANDLER_CONSUMED, 404 RX_HANDLER_ANOTHER, 405 RX_HANDLER_EXACT, 406 RX_HANDLER_PASS, 407 }; 408 typedef enum rx_handler_result rx_handler_result_t; 409 typedef rx_handler_result_t rx_handler_func_t(struct sk_buff **pskb); 410 411 void __napi_schedule(struct napi_struct *n); 412 void __napi_schedule_irqoff(struct napi_struct *n); 413 414 static inline bool napi_disable_pending(struct napi_struct *n) 415 { 416 return test_bit(NAPI_STATE_DISABLE, &n->state); 417 } 418 419 bool napi_schedule_prep(struct napi_struct *n); 420 421 /** 422 * napi_schedule - schedule NAPI poll 423 * @n: NAPI context 424 * 425 * Schedule NAPI poll routine to be called if it is not already 426 * running. 427 */ 428 static inline void napi_schedule(struct napi_struct *n) 429 { 430 if (napi_schedule_prep(n)) 431 __napi_schedule(n); 432 } 433 434 /** 435 * napi_schedule_irqoff - schedule NAPI poll 436 * @n: NAPI context 437 * 438 * Variant of napi_schedule(), assuming hard irqs are masked. 439 */ 440 static inline void napi_schedule_irqoff(struct napi_struct *n) 441 { 442 if (napi_schedule_prep(n)) 443 __napi_schedule_irqoff(n); 444 } 445 446 /* Try to reschedule poll. Called by dev->poll() after napi_complete(). */ 447 static inline bool napi_reschedule(struct napi_struct *napi) 448 { 449 if (napi_schedule_prep(napi)) { 450 __napi_schedule(napi); 451 return true; 452 } 453 return false; 454 } 455 456 bool napi_complete_done(struct napi_struct *n, int work_done); 457 /** 458 * napi_complete - NAPI processing complete 459 * @n: NAPI context 460 * 461 * Mark NAPI processing as complete. 462 * Consider using napi_complete_done() instead. 463 * Return false if device should avoid rearming interrupts. 464 */ 465 static inline bool napi_complete(struct napi_struct *n) 466 { 467 return napi_complete_done(n, 0); 468 } 469 470 /** 471 * napi_hash_del - remove a NAPI from global table 472 * @napi: NAPI context 473 * 474 * Warning: caller must observe RCU grace period 475 * before freeing memory containing @napi, if 476 * this function returns true. 477 * Note: core networking stack automatically calls it 478 * from netif_napi_del(). 479 * Drivers might want to call this helper to combine all 480 * the needed RCU grace periods into a single one. 481 */ 482 bool napi_hash_del(struct napi_struct *napi); 483 484 /** 485 * napi_disable - prevent NAPI from scheduling 486 * @n: NAPI context 487 * 488 * Stop NAPI from being scheduled on this context. 489 * Waits till any outstanding processing completes. 490 */ 491 void napi_disable(struct napi_struct *n); 492 493 /** 494 * napi_enable - enable NAPI scheduling 495 * @n: NAPI context 496 * 497 * Resume NAPI from being scheduled on this context. 498 * Must be paired with napi_disable. 499 */ 500 static inline void napi_enable(struct napi_struct *n) 501 { 502 BUG_ON(!test_bit(NAPI_STATE_SCHED, &n->state)); 503 smp_mb__before_atomic(); 504 clear_bit(NAPI_STATE_SCHED, &n->state); 505 clear_bit(NAPI_STATE_NPSVC, &n->state); 506 } 507 508 /** 509 * napi_synchronize - wait until NAPI is not running 510 * @n: NAPI context 511 * 512 * Wait until NAPI is done being scheduled on this context. 513 * Waits till any outstanding processing completes but 514 * does not disable future activations. 515 */ 516 static inline void napi_synchronize(const struct napi_struct *n) 517 { 518 if (IS_ENABLED(CONFIG_SMP)) 519 while (test_bit(NAPI_STATE_SCHED, &n->state)) 520 msleep(1); 521 else 522 barrier(); 523 } 524 525 enum netdev_queue_state_t { 526 __QUEUE_STATE_DRV_XOFF, 527 __QUEUE_STATE_STACK_XOFF, 528 __QUEUE_STATE_FROZEN, 529 }; 530 531 #define QUEUE_STATE_DRV_XOFF (1 << __QUEUE_STATE_DRV_XOFF) 532 #define QUEUE_STATE_STACK_XOFF (1 << __QUEUE_STATE_STACK_XOFF) 533 #define QUEUE_STATE_FROZEN (1 << __QUEUE_STATE_FROZEN) 534 535 #define QUEUE_STATE_ANY_XOFF (QUEUE_STATE_DRV_XOFF | QUEUE_STATE_STACK_XOFF) 536 #define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \ 537 QUEUE_STATE_FROZEN) 538 #define QUEUE_STATE_DRV_XOFF_OR_FROZEN (QUEUE_STATE_DRV_XOFF | \ 539 QUEUE_STATE_FROZEN) 540 541 /* 542 * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue. The 543 * netif_tx_* functions below are used to manipulate this flag. The 544 * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit 545 * queue independently. The netif_xmit_*stopped functions below are called 546 * to check if the queue has been stopped by the driver or stack (either 547 * of the XOFF bits are set in the state). Drivers should not need to call 548 * netif_xmit*stopped functions, they should only be using netif_tx_*. 549 */ 550 551 struct netdev_queue { 552 /* 553 * read-mostly part 554 */ 555 struct net_device *dev; 556 struct Qdisc __rcu *qdisc; 557 struct Qdisc *qdisc_sleeping; 558 #ifdef CONFIG_SYSFS 559 struct kobject kobj; 560 #endif 561 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA) 562 int numa_node; 563 #endif 564 unsigned long tx_maxrate; 565 /* 566 * Number of TX timeouts for this queue 567 * (/sys/class/net/DEV/Q/trans_timeout) 568 */ 569 unsigned long trans_timeout; 570 /* 571 * write-mostly part 572 */ 573 spinlock_t _xmit_lock ____cacheline_aligned_in_smp; 574 int xmit_lock_owner; 575 /* 576 * Time (in jiffies) of last Tx 577 */ 578 unsigned long trans_start; 579 580 unsigned long state; 581 582 #ifdef CONFIG_BQL 583 struct dql dql; 584 #endif 585 } ____cacheline_aligned_in_smp; 586 587 static inline int netdev_queue_numa_node_read(const struct netdev_queue *q) 588 { 589 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA) 590 return q->numa_node; 591 #else 592 return NUMA_NO_NODE; 593 #endif 594 } 595 596 static inline void netdev_queue_numa_node_write(struct netdev_queue *q, int node) 597 { 598 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA) 599 q->numa_node = node; 600 #endif 601 } 602 603 #ifdef CONFIG_RPS 604 /* 605 * This structure holds an RPS map which can be of variable length. The 606 * map is an array of CPUs. 607 */ 608 struct rps_map { 609 unsigned int len; 610 struct rcu_head rcu; 611 u16 cpus[0]; 612 }; 613 #define RPS_MAP_SIZE(_num) (sizeof(struct rps_map) + ((_num) * sizeof(u16))) 614 615 /* 616 * The rps_dev_flow structure contains the mapping of a flow to a CPU, the 617 * tail pointer for that CPU's input queue at the time of last enqueue, and 618 * a hardware filter index. 619 */ 620 struct rps_dev_flow { 621 u16 cpu; 622 u16 filter; 623 unsigned int last_qtail; 624 }; 625 #define RPS_NO_FILTER 0xffff 626 627 /* 628 * The rps_dev_flow_table structure contains a table of flow mappings. 629 */ 630 struct rps_dev_flow_table { 631 unsigned int mask; 632 struct rcu_head rcu; 633 struct rps_dev_flow flows[0]; 634 }; 635 #define RPS_DEV_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_dev_flow_table) + \ 636 ((_num) * sizeof(struct rps_dev_flow))) 637 638 /* 639 * The rps_sock_flow_table contains mappings of flows to the last CPU 640 * on which they were processed by the application (set in recvmsg). 641 * Each entry is a 32bit value. Upper part is the high-order bits 642 * of flow hash, lower part is CPU number. 643 * rps_cpu_mask is used to partition the space, depending on number of 644 * possible CPUs : rps_cpu_mask = roundup_pow_of_two(nr_cpu_ids) - 1 645 * For example, if 64 CPUs are possible, rps_cpu_mask = 0x3f, 646 * meaning we use 32-6=26 bits for the hash. 647 */ 648 struct rps_sock_flow_table { 649 u32 mask; 650 651 u32 ents[0] ____cacheline_aligned_in_smp; 652 }; 653 #define RPS_SOCK_FLOW_TABLE_SIZE(_num) (offsetof(struct rps_sock_flow_table, ents[_num])) 654 655 #define RPS_NO_CPU 0xffff 656 657 extern u32 rps_cpu_mask; 658 extern struct rps_sock_flow_table __rcu *rps_sock_flow_table; 659 660 static inline void rps_record_sock_flow(struct rps_sock_flow_table *table, 661 u32 hash) 662 { 663 if (table && hash) { 664 unsigned int index = hash & table->mask; 665 u32 val = hash & ~rps_cpu_mask; 666 667 /* We only give a hint, preemption can change CPU under us */ 668 val |= raw_smp_processor_id(); 669 670 if (table->ents[index] != val) 671 table->ents[index] = val; 672 } 673 } 674 675 #ifdef CONFIG_RFS_ACCEL 676 bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index, u32 flow_id, 677 u16 filter_id); 678 #endif 679 #endif /* CONFIG_RPS */ 680 681 /* This structure contains an instance of an RX queue. */ 682 struct netdev_rx_queue { 683 #ifdef CONFIG_RPS 684 struct rps_map __rcu *rps_map; 685 struct rps_dev_flow_table __rcu *rps_flow_table; 686 #endif 687 struct kobject kobj; 688 struct net_device *dev; 689 } ____cacheline_aligned_in_smp; 690 691 /* 692 * RX queue sysfs structures and functions. 693 */ 694 struct rx_queue_attribute { 695 struct attribute attr; 696 ssize_t (*show)(struct netdev_rx_queue *queue, char *buf); 697 ssize_t (*store)(struct netdev_rx_queue *queue, 698 const char *buf, size_t len); 699 }; 700 701 #ifdef CONFIG_XPS 702 /* 703 * This structure holds an XPS map which can be of variable length. The 704 * map is an array of queues. 705 */ 706 struct xps_map { 707 unsigned int len; 708 unsigned int alloc_len; 709 struct rcu_head rcu; 710 u16 queues[0]; 711 }; 712 #define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16))) 713 #define XPS_MIN_MAP_ALLOC ((L1_CACHE_ALIGN(offsetof(struct xps_map, queues[1])) \ 714 - sizeof(struct xps_map)) / sizeof(u16)) 715 716 /* 717 * This structure holds all XPS maps for device. Maps are indexed by CPU. 718 */ 719 struct xps_dev_maps { 720 struct rcu_head rcu; 721 struct xps_map __rcu *cpu_map[0]; 722 }; 723 #define XPS_DEV_MAPS_SIZE(_tcs) (sizeof(struct xps_dev_maps) + \ 724 (nr_cpu_ids * (_tcs) * sizeof(struct xps_map *))) 725 #endif /* CONFIG_XPS */ 726 727 #define TC_MAX_QUEUE 16 728 #define TC_BITMASK 15 729 /* HW offloaded queuing disciplines txq count and offset maps */ 730 struct netdev_tc_txq { 731 u16 count; 732 u16 offset; 733 }; 734 735 #if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE) 736 /* 737 * This structure is to hold information about the device 738 * configured to run FCoE protocol stack. 739 */ 740 struct netdev_fcoe_hbainfo { 741 char manufacturer[64]; 742 char serial_number[64]; 743 char hardware_version[64]; 744 char driver_version[64]; 745 char optionrom_version[64]; 746 char firmware_version[64]; 747 char model[256]; 748 char model_description[256]; 749 }; 750 #endif 751 752 #define MAX_PHYS_ITEM_ID_LEN 32 753 754 /* This structure holds a unique identifier to identify some 755 * physical item (port for example) used by a netdevice. 756 */ 757 struct netdev_phys_item_id { 758 unsigned char id[MAX_PHYS_ITEM_ID_LEN]; 759 unsigned char id_len; 760 }; 761 762 static inline bool netdev_phys_item_id_same(struct netdev_phys_item_id *a, 763 struct netdev_phys_item_id *b) 764 { 765 return a->id_len == b->id_len && 766 memcmp(a->id, b->id, a->id_len) == 0; 767 } 768 769 typedef u16 (*select_queue_fallback_t)(struct net_device *dev, 770 struct sk_buff *skb); 771 772 enum tc_setup_type { 773 TC_SETUP_MQPRIO, 774 TC_SETUP_CLSU32, 775 TC_SETUP_CLSFLOWER, 776 TC_SETUP_CLSMATCHALL, 777 TC_SETUP_CLSBPF, 778 }; 779 780 /* These structures hold the attributes of xdp state that are being passed 781 * to the netdevice through the xdp op. 782 */ 783 enum xdp_netdev_command { 784 /* Set or clear a bpf program used in the earliest stages of packet 785 * rx. The prog will have been loaded as BPF_PROG_TYPE_XDP. The callee 786 * is responsible for calling bpf_prog_put on any old progs that are 787 * stored. In case of error, the callee need not release the new prog 788 * reference, but on success it takes ownership and must bpf_prog_put 789 * when it is no longer used. 790 */ 791 XDP_SETUP_PROG, 792 XDP_SETUP_PROG_HW, 793 /* Check if a bpf program is set on the device. The callee should 794 * set @prog_attached to one of XDP_ATTACHED_* values, note that "true" 795 * is equivalent to XDP_ATTACHED_DRV. 796 */ 797 XDP_QUERY_PROG, 798 }; 799 800 struct netlink_ext_ack; 801 802 struct netdev_xdp { 803 enum xdp_netdev_command command; 804 union { 805 /* XDP_SETUP_PROG */ 806 struct { 807 u32 flags; 808 struct bpf_prog *prog; 809 struct netlink_ext_ack *extack; 810 }; 811 /* XDP_QUERY_PROG */ 812 struct { 813 u8 prog_attached; 814 u32 prog_id; 815 }; 816 }; 817 }; 818 819 #ifdef CONFIG_XFRM_OFFLOAD 820 struct xfrmdev_ops { 821 int (*xdo_dev_state_add) (struct xfrm_state *x); 822 void (*xdo_dev_state_delete) (struct xfrm_state *x); 823 void (*xdo_dev_state_free) (struct xfrm_state *x); 824 bool (*xdo_dev_offload_ok) (struct sk_buff *skb, 825 struct xfrm_state *x); 826 }; 827 #endif 828 829 struct dev_ifalias { 830 struct rcu_head rcuhead; 831 char ifalias[]; 832 }; 833 834 /* 835 * This structure defines the management hooks for network devices. 836 * The following hooks can be defined; unless noted otherwise, they are 837 * optional and can be filled with a null pointer. 838 * 839 * int (*ndo_init)(struct net_device *dev); 840 * This function is called once when a network device is registered. 841 * The network device can use this for any late stage initialization 842 * or semantic validation. It can fail with an error code which will 843 * be propagated back to register_netdev. 844 * 845 * void (*ndo_uninit)(struct net_device *dev); 846 * This function is called when device is unregistered or when registration 847 * fails. It is not called if init fails. 848 * 849 * int (*ndo_open)(struct net_device *dev); 850 * This function is called when a network device transitions to the up 851 * state. 852 * 853 * int (*ndo_stop)(struct net_device *dev); 854 * This function is called when a network device transitions to the down 855 * state. 856 * 857 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb, 858 * struct net_device *dev); 859 * Called when a packet needs to be transmitted. 860 * Returns NETDEV_TX_OK. Can return NETDEV_TX_BUSY, but you should stop 861 * the queue before that can happen; it's for obsolete devices and weird 862 * corner cases, but the stack really does a non-trivial amount 863 * of useless work if you return NETDEV_TX_BUSY. 864 * Required; cannot be NULL. 865 * 866 * netdev_features_t (*ndo_features_check)(struct sk_buff *skb, 867 * struct net_device *dev 868 * netdev_features_t features); 869 * Called by core transmit path to determine if device is capable of 870 * performing offload operations on a given packet. This is to give 871 * the device an opportunity to implement any restrictions that cannot 872 * be otherwise expressed by feature flags. The check is called with 873 * the set of features that the stack has calculated and it returns 874 * those the driver believes to be appropriate. 875 * 876 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb, 877 * void *accel_priv, select_queue_fallback_t fallback); 878 * Called to decide which queue to use when device supports multiple 879 * transmit queues. 880 * 881 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags); 882 * This function is called to allow device receiver to make 883 * changes to configuration when multicast or promiscuous is enabled. 884 * 885 * void (*ndo_set_rx_mode)(struct net_device *dev); 886 * This function is called device changes address list filtering. 887 * If driver handles unicast address filtering, it should set 888 * IFF_UNICAST_FLT in its priv_flags. 889 * 890 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr); 891 * This function is called when the Media Access Control address 892 * needs to be changed. If this interface is not defined, the 893 * MAC address can not be changed. 894 * 895 * int (*ndo_validate_addr)(struct net_device *dev); 896 * Test if Media Access Control address is valid for the device. 897 * 898 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd); 899 * Called when a user requests an ioctl which can't be handled by 900 * the generic interface code. If not defined ioctls return 901 * not supported error code. 902 * 903 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map); 904 * Used to set network devices bus interface parameters. This interface 905 * is retained for legacy reasons; new devices should use the bus 906 * interface (PCI) for low level management. 907 * 908 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu); 909 * Called when a user wants to change the Maximum Transfer Unit 910 * of a device. 911 * 912 * void (*ndo_tx_timeout)(struct net_device *dev); 913 * Callback used when the transmitter has not made any progress 914 * for dev->watchdog ticks. 915 * 916 * void (*ndo_get_stats64)(struct net_device *dev, 917 * struct rtnl_link_stats64 *storage); 918 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev); 919 * Called when a user wants to get the network device usage 920 * statistics. Drivers must do one of the following: 921 * 1. Define @ndo_get_stats64 to fill in a zero-initialised 922 * rtnl_link_stats64 structure passed by the caller. 923 * 2. Define @ndo_get_stats to update a net_device_stats structure 924 * (which should normally be dev->stats) and return a pointer to 925 * it. The structure may be changed asynchronously only if each 926 * field is written atomically. 927 * 3. Update dev->stats asynchronously and atomically, and define 928 * neither operation. 929 * 930 * bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id) 931 * Return true if this device supports offload stats of this attr_id. 932 * 933 * int (*ndo_get_offload_stats)(int attr_id, const struct net_device *dev, 934 * void *attr_data) 935 * Get statistics for offload operations by attr_id. Write it into the 936 * attr_data pointer. 937 * 938 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16 vid); 939 * If device supports VLAN filtering this function is called when a 940 * VLAN id is registered. 941 * 942 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, __be16 proto, u16 vid); 943 * If device supports VLAN filtering this function is called when a 944 * VLAN id is unregistered. 945 * 946 * void (*ndo_poll_controller)(struct net_device *dev); 947 * 948 * SR-IOV management functions. 949 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac); 950 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan, 951 * u8 qos, __be16 proto); 952 * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate, 953 * int max_tx_rate); 954 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting); 955 * int (*ndo_set_vf_trust)(struct net_device *dev, int vf, bool setting); 956 * int (*ndo_get_vf_config)(struct net_device *dev, 957 * int vf, struct ifla_vf_info *ivf); 958 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state); 959 * int (*ndo_set_vf_port)(struct net_device *dev, int vf, 960 * struct nlattr *port[]); 961 * 962 * Enable or disable the VF ability to query its RSS Redirection Table and 963 * Hash Key. This is needed since on some devices VF share this information 964 * with PF and querying it may introduce a theoretical security risk. 965 * int (*ndo_set_vf_rss_query_en)(struct net_device *dev, int vf, bool setting); 966 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb); 967 * int (*ndo_setup_tc)(struct net_device *dev, enum tc_setup_type type, 968 * void *type_data); 969 * Called to setup any 'tc' scheduler, classifier or action on @dev. 970 * This is always called from the stack with the rtnl lock held and netif 971 * tx queues stopped. This allows the netdevice to perform queue 972 * management safely. 973 * 974 * Fiber Channel over Ethernet (FCoE) offload functions. 975 * int (*ndo_fcoe_enable)(struct net_device *dev); 976 * Called when the FCoE protocol stack wants to start using LLD for FCoE 977 * so the underlying device can perform whatever needed configuration or 978 * initialization to support acceleration of FCoE traffic. 979 * 980 * int (*ndo_fcoe_disable)(struct net_device *dev); 981 * Called when the FCoE protocol stack wants to stop using LLD for FCoE 982 * so the underlying device can perform whatever needed clean-ups to 983 * stop supporting acceleration of FCoE traffic. 984 * 985 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid, 986 * struct scatterlist *sgl, unsigned int sgc); 987 * Called when the FCoE Initiator wants to initialize an I/O that 988 * is a possible candidate for Direct Data Placement (DDP). The LLD can 989 * perform necessary setup and returns 1 to indicate the device is set up 990 * successfully to perform DDP on this I/O, otherwise this returns 0. 991 * 992 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid); 993 * Called when the FCoE Initiator/Target is done with the DDPed I/O as 994 * indicated by the FC exchange id 'xid', so the underlying device can 995 * clean up and reuse resources for later DDP requests. 996 * 997 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid, 998 * struct scatterlist *sgl, unsigned int sgc); 999 * Called when the FCoE Target wants to initialize an I/O that 1000 * is a possible candidate for Direct Data Placement (DDP). The LLD can 1001 * perform necessary setup and returns 1 to indicate the device is set up 1002 * successfully to perform DDP on this I/O, otherwise this returns 0. 1003 * 1004 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev, 1005 * struct netdev_fcoe_hbainfo *hbainfo); 1006 * Called when the FCoE Protocol stack wants information on the underlying 1007 * device. This information is utilized by the FCoE protocol stack to 1008 * register attributes with Fiber Channel management service as per the 1009 * FC-GS Fabric Device Management Information(FDMI) specification. 1010 * 1011 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type); 1012 * Called when the underlying device wants to override default World Wide 1013 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own 1014 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE 1015 * protocol stack to use. 1016 * 1017 * RFS acceleration. 1018 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb, 1019 * u16 rxq_index, u32 flow_id); 1020 * Set hardware filter for RFS. rxq_index is the target queue index; 1021 * flow_id is a flow ID to be passed to rps_may_expire_flow() later. 1022 * Return the filter ID on success, or a negative error code. 1023 * 1024 * Slave management functions (for bridge, bonding, etc). 1025 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev); 1026 * Called to make another netdev an underling. 1027 * 1028 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev); 1029 * Called to release previously enslaved netdev. 1030 * 1031 * Feature/offload setting functions. 1032 * netdev_features_t (*ndo_fix_features)(struct net_device *dev, 1033 * netdev_features_t features); 1034 * Adjusts the requested feature flags according to device-specific 1035 * constraints, and returns the resulting flags. Must not modify 1036 * the device state. 1037 * 1038 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features); 1039 * Called to update device configuration to new features. Passed 1040 * feature set might be less than what was returned by ndo_fix_features()). 1041 * Must return >0 or -errno if it changed dev->features itself. 1042 * 1043 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[], 1044 * struct net_device *dev, 1045 * const unsigned char *addr, u16 vid, u16 flags) 1046 * Adds an FDB entry to dev for addr. 1047 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[], 1048 * struct net_device *dev, 1049 * const unsigned char *addr, u16 vid) 1050 * Deletes the FDB entry from dev coresponding to addr. 1051 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb, 1052 * struct net_device *dev, struct net_device *filter_dev, 1053 * int *idx) 1054 * Used to add FDB entries to dump requests. Implementers should add 1055 * entries to skb and update idx with the number of entries. 1056 * 1057 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh, 1058 * u16 flags) 1059 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq, 1060 * struct net_device *dev, u32 filter_mask, 1061 * int nlflags) 1062 * int (*ndo_bridge_dellink)(struct net_device *dev, struct nlmsghdr *nlh, 1063 * u16 flags); 1064 * 1065 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier); 1066 * Called to change device carrier. Soft-devices (like dummy, team, etc) 1067 * which do not represent real hardware may define this to allow their 1068 * userspace components to manage their virtual carrier state. Devices 1069 * that determine carrier state from physical hardware properties (eg 1070 * network cables) or protocol-dependent mechanisms (eg 1071 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function. 1072 * 1073 * int (*ndo_get_phys_port_id)(struct net_device *dev, 1074 * struct netdev_phys_item_id *ppid); 1075 * Called to get ID of physical port of this device. If driver does 1076 * not implement this, it is assumed that the hw is not able to have 1077 * multiple net devices on single physical port. 1078 * 1079 * void (*ndo_udp_tunnel_add)(struct net_device *dev, 1080 * struct udp_tunnel_info *ti); 1081 * Called by UDP tunnel to notify a driver about the UDP port and socket 1082 * address family that a UDP tunnel is listnening to. It is called only 1083 * when a new port starts listening. The operation is protected by the 1084 * RTNL. 1085 * 1086 * void (*ndo_udp_tunnel_del)(struct net_device *dev, 1087 * struct udp_tunnel_info *ti); 1088 * Called by UDP tunnel to notify the driver about a UDP port and socket 1089 * address family that the UDP tunnel is not listening to anymore. The 1090 * operation is protected by the RTNL. 1091 * 1092 * void* (*ndo_dfwd_add_station)(struct net_device *pdev, 1093 * struct net_device *dev) 1094 * Called by upper layer devices to accelerate switching or other 1095 * station functionality into hardware. 'pdev is the lowerdev 1096 * to use for the offload and 'dev' is the net device that will 1097 * back the offload. Returns a pointer to the private structure 1098 * the upper layer will maintain. 1099 * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv) 1100 * Called by upper layer device to delete the station created 1101 * by 'ndo_dfwd_add_station'. 'pdev' is the net device backing 1102 * the station and priv is the structure returned by the add 1103 * operation. 1104 * int (*ndo_set_tx_maxrate)(struct net_device *dev, 1105 * int queue_index, u32 maxrate); 1106 * Called when a user wants to set a max-rate limitation of specific 1107 * TX queue. 1108 * int (*ndo_get_iflink)(const struct net_device *dev); 1109 * Called to get the iflink value of this device. 1110 * void (*ndo_change_proto_down)(struct net_device *dev, 1111 * bool proto_down); 1112 * This function is used to pass protocol port error state information 1113 * to the switch driver. The switch driver can react to the proto_down 1114 * by doing a phys down on the associated switch port. 1115 * int (*ndo_fill_metadata_dst)(struct net_device *dev, struct sk_buff *skb); 1116 * This function is used to get egress tunnel information for given skb. 1117 * This is useful for retrieving outer tunnel header parameters while 1118 * sampling packet. 1119 * void (*ndo_set_rx_headroom)(struct net_device *dev, int needed_headroom); 1120 * This function is used to specify the headroom that the skb must 1121 * consider when allocation skb during packet reception. Setting 1122 * appropriate rx headroom value allows avoiding skb head copy on 1123 * forward. Setting a negative value resets the rx headroom to the 1124 * default value. 1125 * int (*ndo_xdp)(struct net_device *dev, struct netdev_xdp *xdp); 1126 * This function is used to set or query state related to XDP on the 1127 * netdevice. See definition of enum xdp_netdev_command for details. 1128 * int (*ndo_xdp_xmit)(struct net_device *dev, struct xdp_buff *xdp); 1129 * This function is used to submit a XDP packet for transmit on a 1130 * netdevice. 1131 * void (*ndo_xdp_flush)(struct net_device *dev); 1132 * This function is used to inform the driver to flush a particular 1133 * xdp tx queue. Must be called on same CPU as xdp_xmit. 1134 */ 1135 struct net_device_ops { 1136 int (*ndo_init)(struct net_device *dev); 1137 void (*ndo_uninit)(struct net_device *dev); 1138 int (*ndo_open)(struct net_device *dev); 1139 int (*ndo_stop)(struct net_device *dev); 1140 netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb, 1141 struct net_device *dev); 1142 netdev_features_t (*ndo_features_check)(struct sk_buff *skb, 1143 struct net_device *dev, 1144 netdev_features_t features); 1145 u16 (*ndo_select_queue)(struct net_device *dev, 1146 struct sk_buff *skb, 1147 void *accel_priv, 1148 select_queue_fallback_t fallback); 1149 void (*ndo_change_rx_flags)(struct net_device *dev, 1150 int flags); 1151 void (*ndo_set_rx_mode)(struct net_device *dev); 1152 int (*ndo_set_mac_address)(struct net_device *dev, 1153 void *addr); 1154 int (*ndo_validate_addr)(struct net_device *dev); 1155 int (*ndo_do_ioctl)(struct net_device *dev, 1156 struct ifreq *ifr, int cmd); 1157 int (*ndo_set_config)(struct net_device *dev, 1158 struct ifmap *map); 1159 int (*ndo_change_mtu)(struct net_device *dev, 1160 int new_mtu); 1161 int (*ndo_neigh_setup)(struct net_device *dev, 1162 struct neigh_parms *); 1163 void (*ndo_tx_timeout) (struct net_device *dev); 1164 1165 void (*ndo_get_stats64)(struct net_device *dev, 1166 struct rtnl_link_stats64 *storage); 1167 bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id); 1168 int (*ndo_get_offload_stats)(int attr_id, 1169 const struct net_device *dev, 1170 void *attr_data); 1171 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev); 1172 1173 int (*ndo_vlan_rx_add_vid)(struct net_device *dev, 1174 __be16 proto, u16 vid); 1175 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, 1176 __be16 proto, u16 vid); 1177 #ifdef CONFIG_NET_POLL_CONTROLLER 1178 void (*ndo_poll_controller)(struct net_device *dev); 1179 int (*ndo_netpoll_setup)(struct net_device *dev, 1180 struct netpoll_info *info); 1181 void (*ndo_netpoll_cleanup)(struct net_device *dev); 1182 #endif 1183 int (*ndo_set_vf_mac)(struct net_device *dev, 1184 int queue, u8 *mac); 1185 int (*ndo_set_vf_vlan)(struct net_device *dev, 1186 int queue, u16 vlan, 1187 u8 qos, __be16 proto); 1188 int (*ndo_set_vf_rate)(struct net_device *dev, 1189 int vf, int min_tx_rate, 1190 int max_tx_rate); 1191 int (*ndo_set_vf_spoofchk)(struct net_device *dev, 1192 int vf, bool setting); 1193 int (*ndo_set_vf_trust)(struct net_device *dev, 1194 int vf, bool setting); 1195 int (*ndo_get_vf_config)(struct net_device *dev, 1196 int vf, 1197 struct ifla_vf_info *ivf); 1198 int (*ndo_set_vf_link_state)(struct net_device *dev, 1199 int vf, int link_state); 1200 int (*ndo_get_vf_stats)(struct net_device *dev, 1201 int vf, 1202 struct ifla_vf_stats 1203 *vf_stats); 1204 int (*ndo_set_vf_port)(struct net_device *dev, 1205 int vf, 1206 struct nlattr *port[]); 1207 int (*ndo_get_vf_port)(struct net_device *dev, 1208 int vf, struct sk_buff *skb); 1209 int (*ndo_set_vf_guid)(struct net_device *dev, 1210 int vf, u64 guid, 1211 int guid_type); 1212 int (*ndo_set_vf_rss_query_en)( 1213 struct net_device *dev, 1214 int vf, bool setting); 1215 int (*ndo_setup_tc)(struct net_device *dev, 1216 enum tc_setup_type type, 1217 void *type_data); 1218 #if IS_ENABLED(CONFIG_FCOE) 1219 int (*ndo_fcoe_enable)(struct net_device *dev); 1220 int (*ndo_fcoe_disable)(struct net_device *dev); 1221 int (*ndo_fcoe_ddp_setup)(struct net_device *dev, 1222 u16 xid, 1223 struct scatterlist *sgl, 1224 unsigned int sgc); 1225 int (*ndo_fcoe_ddp_done)(struct net_device *dev, 1226 u16 xid); 1227 int (*ndo_fcoe_ddp_target)(struct net_device *dev, 1228 u16 xid, 1229 struct scatterlist *sgl, 1230 unsigned int sgc); 1231 int (*ndo_fcoe_get_hbainfo)(struct net_device *dev, 1232 struct netdev_fcoe_hbainfo *hbainfo); 1233 #endif 1234 1235 #if IS_ENABLED(CONFIG_LIBFCOE) 1236 #define NETDEV_FCOE_WWNN 0 1237 #define NETDEV_FCOE_WWPN 1 1238 int (*ndo_fcoe_get_wwn)(struct net_device *dev, 1239 u64 *wwn, int type); 1240 #endif 1241 1242 #ifdef CONFIG_RFS_ACCEL 1243 int (*ndo_rx_flow_steer)(struct net_device *dev, 1244 const struct sk_buff *skb, 1245 u16 rxq_index, 1246 u32 flow_id); 1247 #endif 1248 int (*ndo_add_slave)(struct net_device *dev, 1249 struct net_device *slave_dev); 1250 int (*ndo_del_slave)(struct net_device *dev, 1251 struct net_device *slave_dev); 1252 netdev_features_t (*ndo_fix_features)(struct net_device *dev, 1253 netdev_features_t features); 1254 int (*ndo_set_features)(struct net_device *dev, 1255 netdev_features_t features); 1256 int (*ndo_neigh_construct)(struct net_device *dev, 1257 struct neighbour *n); 1258 void (*ndo_neigh_destroy)(struct net_device *dev, 1259 struct neighbour *n); 1260 1261 int (*ndo_fdb_add)(struct ndmsg *ndm, 1262 struct nlattr *tb[], 1263 struct net_device *dev, 1264 const unsigned char *addr, 1265 u16 vid, 1266 u16 flags); 1267 int (*ndo_fdb_del)(struct ndmsg *ndm, 1268 struct nlattr *tb[], 1269 struct net_device *dev, 1270 const unsigned char *addr, 1271 u16 vid); 1272 int (*ndo_fdb_dump)(struct sk_buff *skb, 1273 struct netlink_callback *cb, 1274 struct net_device *dev, 1275 struct net_device *filter_dev, 1276 int *idx); 1277 1278 int (*ndo_bridge_setlink)(struct net_device *dev, 1279 struct nlmsghdr *nlh, 1280 u16 flags); 1281 int (*ndo_bridge_getlink)(struct sk_buff *skb, 1282 u32 pid, u32 seq, 1283 struct net_device *dev, 1284 u32 filter_mask, 1285 int nlflags); 1286 int (*ndo_bridge_dellink)(struct net_device *dev, 1287 struct nlmsghdr *nlh, 1288 u16 flags); 1289 int (*ndo_change_carrier)(struct net_device *dev, 1290 bool new_carrier); 1291 int (*ndo_get_phys_port_id)(struct net_device *dev, 1292 struct netdev_phys_item_id *ppid); 1293 int (*ndo_get_phys_port_name)(struct net_device *dev, 1294 char *name, size_t len); 1295 void (*ndo_udp_tunnel_add)(struct net_device *dev, 1296 struct udp_tunnel_info *ti); 1297 void (*ndo_udp_tunnel_del)(struct net_device *dev, 1298 struct udp_tunnel_info *ti); 1299 void* (*ndo_dfwd_add_station)(struct net_device *pdev, 1300 struct net_device *dev); 1301 void (*ndo_dfwd_del_station)(struct net_device *pdev, 1302 void *priv); 1303 1304 int (*ndo_get_lock_subclass)(struct net_device *dev); 1305 int (*ndo_set_tx_maxrate)(struct net_device *dev, 1306 int queue_index, 1307 u32 maxrate); 1308 int (*ndo_get_iflink)(const struct net_device *dev); 1309 int (*ndo_change_proto_down)(struct net_device *dev, 1310 bool proto_down); 1311 int (*ndo_fill_metadata_dst)(struct net_device *dev, 1312 struct sk_buff *skb); 1313 void (*ndo_set_rx_headroom)(struct net_device *dev, 1314 int needed_headroom); 1315 int (*ndo_xdp)(struct net_device *dev, 1316 struct netdev_xdp *xdp); 1317 int (*ndo_xdp_xmit)(struct net_device *dev, 1318 struct xdp_buff *xdp); 1319 void (*ndo_xdp_flush)(struct net_device *dev); 1320 }; 1321 1322 /** 1323 * enum net_device_priv_flags - &struct net_device priv_flags 1324 * 1325 * These are the &struct net_device, they are only set internally 1326 * by drivers and used in the kernel. These flags are invisible to 1327 * userspace; this means that the order of these flags can change 1328 * during any kernel release. 1329 * 1330 * You should have a pretty good reason to be extending these flags. 1331 * 1332 * @IFF_802_1Q_VLAN: 802.1Q VLAN device 1333 * @IFF_EBRIDGE: Ethernet bridging device 1334 * @IFF_BONDING: bonding master or slave 1335 * @IFF_ISATAP: ISATAP interface (RFC4214) 1336 * @IFF_WAN_HDLC: WAN HDLC device 1337 * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to 1338 * release skb->dst 1339 * @IFF_DONT_BRIDGE: disallow bridging this ether dev 1340 * @IFF_DISABLE_NETPOLL: disable netpoll at run-time 1341 * @IFF_MACVLAN_PORT: device used as macvlan port 1342 * @IFF_BRIDGE_PORT: device used as bridge port 1343 * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port 1344 * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit 1345 * @IFF_UNICAST_FLT: Supports unicast filtering 1346 * @IFF_TEAM_PORT: device used as team port 1347 * @IFF_SUPP_NOFCS: device supports sending custom FCS 1348 * @IFF_LIVE_ADDR_CHANGE: device supports hardware address 1349 * change when it's running 1350 * @IFF_MACVLAN: Macvlan device 1351 * @IFF_XMIT_DST_RELEASE_PERM: IFF_XMIT_DST_RELEASE not taking into account 1352 * underlying stacked devices 1353 * @IFF_IPVLAN_MASTER: IPvlan master device 1354 * @IFF_IPVLAN_SLAVE: IPvlan slave device 1355 * @IFF_L3MDEV_MASTER: device is an L3 master device 1356 * @IFF_NO_QUEUE: device can run without qdisc attached 1357 * @IFF_OPENVSWITCH: device is a Open vSwitch master 1358 * @IFF_L3MDEV_SLAVE: device is enslaved to an L3 master device 1359 * @IFF_TEAM: device is a team device 1360 * @IFF_RXFH_CONFIGURED: device has had Rx Flow indirection table configured 1361 * @IFF_PHONY_HEADROOM: the headroom value is controlled by an external 1362 * entity (i.e. the master device for bridged veth) 1363 * @IFF_MACSEC: device is a MACsec device 1364 */ 1365 enum netdev_priv_flags { 1366 IFF_802_1Q_VLAN = 1<<0, 1367 IFF_EBRIDGE = 1<<1, 1368 IFF_BONDING = 1<<2, 1369 IFF_ISATAP = 1<<3, 1370 IFF_WAN_HDLC = 1<<4, 1371 IFF_XMIT_DST_RELEASE = 1<<5, 1372 IFF_DONT_BRIDGE = 1<<6, 1373 IFF_DISABLE_NETPOLL = 1<<7, 1374 IFF_MACVLAN_PORT = 1<<8, 1375 IFF_BRIDGE_PORT = 1<<9, 1376 IFF_OVS_DATAPATH = 1<<10, 1377 IFF_TX_SKB_SHARING = 1<<11, 1378 IFF_UNICAST_FLT = 1<<12, 1379 IFF_TEAM_PORT = 1<<13, 1380 IFF_SUPP_NOFCS = 1<<14, 1381 IFF_LIVE_ADDR_CHANGE = 1<<15, 1382 IFF_MACVLAN = 1<<16, 1383 IFF_XMIT_DST_RELEASE_PERM = 1<<17, 1384 IFF_IPVLAN_MASTER = 1<<18, 1385 IFF_IPVLAN_SLAVE = 1<<19, 1386 IFF_L3MDEV_MASTER = 1<<20, 1387 IFF_NO_QUEUE = 1<<21, 1388 IFF_OPENVSWITCH = 1<<22, 1389 IFF_L3MDEV_SLAVE = 1<<23, 1390 IFF_TEAM = 1<<24, 1391 IFF_RXFH_CONFIGURED = 1<<25, 1392 IFF_PHONY_HEADROOM = 1<<26, 1393 IFF_MACSEC = 1<<27, 1394 }; 1395 1396 #define IFF_802_1Q_VLAN IFF_802_1Q_VLAN 1397 #define IFF_EBRIDGE IFF_EBRIDGE 1398 #define IFF_BONDING IFF_BONDING 1399 #define IFF_ISATAP IFF_ISATAP 1400 #define IFF_WAN_HDLC IFF_WAN_HDLC 1401 #define IFF_XMIT_DST_RELEASE IFF_XMIT_DST_RELEASE 1402 #define IFF_DONT_BRIDGE IFF_DONT_BRIDGE 1403 #define IFF_DISABLE_NETPOLL IFF_DISABLE_NETPOLL 1404 #define IFF_MACVLAN_PORT IFF_MACVLAN_PORT 1405 #define IFF_BRIDGE_PORT IFF_BRIDGE_PORT 1406 #define IFF_OVS_DATAPATH IFF_OVS_DATAPATH 1407 #define IFF_TX_SKB_SHARING IFF_TX_SKB_SHARING 1408 #define IFF_UNICAST_FLT IFF_UNICAST_FLT 1409 #define IFF_TEAM_PORT IFF_TEAM_PORT 1410 #define IFF_SUPP_NOFCS IFF_SUPP_NOFCS 1411 #define IFF_LIVE_ADDR_CHANGE IFF_LIVE_ADDR_CHANGE 1412 #define IFF_MACVLAN IFF_MACVLAN 1413 #define IFF_XMIT_DST_RELEASE_PERM IFF_XMIT_DST_RELEASE_PERM 1414 #define IFF_IPVLAN_MASTER IFF_IPVLAN_MASTER 1415 #define IFF_IPVLAN_SLAVE IFF_IPVLAN_SLAVE 1416 #define IFF_L3MDEV_MASTER IFF_L3MDEV_MASTER 1417 #define IFF_NO_QUEUE IFF_NO_QUEUE 1418 #define IFF_OPENVSWITCH IFF_OPENVSWITCH 1419 #define IFF_L3MDEV_SLAVE IFF_L3MDEV_SLAVE 1420 #define IFF_TEAM IFF_TEAM 1421 #define IFF_RXFH_CONFIGURED IFF_RXFH_CONFIGURED 1422 #define IFF_MACSEC IFF_MACSEC 1423 1424 /** 1425 * struct net_device - The DEVICE structure. 1426 * 1427 * Actually, this whole structure is a big mistake. It mixes I/O 1428 * data with strictly "high-level" data, and it has to know about 1429 * almost every data structure used in the INET module. 1430 * 1431 * @name: This is the first field of the "visible" part of this structure 1432 * (i.e. as seen by users in the "Space.c" file). It is the name 1433 * of the interface. 1434 * 1435 * @name_hlist: Device name hash chain, please keep it close to name[] 1436 * @ifalias: SNMP alias 1437 * @mem_end: Shared memory end 1438 * @mem_start: Shared memory start 1439 * @base_addr: Device I/O address 1440 * @irq: Device IRQ number 1441 * 1442 * @carrier_changes: Stats to monitor carrier on<->off transitions 1443 * 1444 * @state: Generic network queuing layer state, see netdev_state_t 1445 * @dev_list: The global list of network devices 1446 * @napi_list: List entry used for polling NAPI devices 1447 * @unreg_list: List entry when we are unregistering the 1448 * device; see the function unregister_netdev 1449 * @close_list: List entry used when we are closing the device 1450 * @ptype_all: Device-specific packet handlers for all protocols 1451 * @ptype_specific: Device-specific, protocol-specific packet handlers 1452 * 1453 * @adj_list: Directly linked devices, like slaves for bonding 1454 * @features: Currently active device features 1455 * @hw_features: User-changeable features 1456 * 1457 * @wanted_features: User-requested features 1458 * @vlan_features: Mask of features inheritable by VLAN devices 1459 * 1460 * @hw_enc_features: Mask of features inherited by encapsulating devices 1461 * This field indicates what encapsulation 1462 * offloads the hardware is capable of doing, 1463 * and drivers will need to set them appropriately. 1464 * 1465 * @mpls_features: Mask of features inheritable by MPLS 1466 * 1467 * @ifindex: interface index 1468 * @group: The group the device belongs to 1469 * 1470 * @stats: Statistics struct, which was left as a legacy, use 1471 * rtnl_link_stats64 instead 1472 * 1473 * @rx_dropped: Dropped packets by core network, 1474 * do not use this in drivers 1475 * @tx_dropped: Dropped packets by core network, 1476 * do not use this in drivers 1477 * @rx_nohandler: nohandler dropped packets by core network on 1478 * inactive devices, do not use this in drivers 1479 * 1480 * @wireless_handlers: List of functions to handle Wireless Extensions, 1481 * instead of ioctl, 1482 * see <net/iw_handler.h> for details. 1483 * @wireless_data: Instance data managed by the core of wireless extensions 1484 * 1485 * @netdev_ops: Includes several pointers to callbacks, 1486 * if one wants to override the ndo_*() functions 1487 * @ethtool_ops: Management operations 1488 * @ndisc_ops: Includes callbacks for different IPv6 neighbour 1489 * discovery handling. Necessary for e.g. 6LoWPAN. 1490 * @header_ops: Includes callbacks for creating,parsing,caching,etc 1491 * of Layer 2 headers. 1492 * 1493 * @flags: Interface flags (a la BSD) 1494 * @priv_flags: Like 'flags' but invisible to userspace, 1495 * see if.h for the definitions 1496 * @gflags: Global flags ( kept as legacy ) 1497 * @padded: How much padding added by alloc_netdev() 1498 * @operstate: RFC2863 operstate 1499 * @link_mode: Mapping policy to operstate 1500 * @if_port: Selectable AUI, TP, ... 1501 * @dma: DMA channel 1502 * @mtu: Interface MTU value 1503 * @min_mtu: Interface Minimum MTU value 1504 * @max_mtu: Interface Maximum MTU value 1505 * @type: Interface hardware type 1506 * @hard_header_len: Maximum hardware header length. 1507 * @min_header_len: Minimum hardware header length 1508 * 1509 * @needed_headroom: Extra headroom the hardware may need, but not in all 1510 * cases can this be guaranteed 1511 * @needed_tailroom: Extra tailroom the hardware may need, but not in all 1512 * cases can this be guaranteed. Some cases also use 1513 * LL_MAX_HEADER instead to allocate the skb 1514 * 1515 * interface address info: 1516 * 1517 * @perm_addr: Permanent hw address 1518 * @addr_assign_type: Hw address assignment type 1519 * @addr_len: Hardware address length 1520 * @neigh_priv_len: Used in neigh_alloc() 1521 * @dev_id: Used to differentiate devices that share 1522 * the same link layer address 1523 * @dev_port: Used to differentiate devices that share 1524 * the same function 1525 * @addr_list_lock: XXX: need comments on this one 1526 * @uc_promisc: Counter that indicates promiscuous mode 1527 * has been enabled due to the need to listen to 1528 * additional unicast addresses in a device that 1529 * does not implement ndo_set_rx_mode() 1530 * @uc: unicast mac addresses 1531 * @mc: multicast mac addresses 1532 * @dev_addrs: list of device hw addresses 1533 * @queues_kset: Group of all Kobjects in the Tx and RX queues 1534 * @promiscuity: Number of times the NIC is told to work in 1535 * promiscuous mode; if it becomes 0 the NIC will 1536 * exit promiscuous mode 1537 * @allmulti: Counter, enables or disables allmulticast mode 1538 * 1539 * @vlan_info: VLAN info 1540 * @dsa_ptr: dsa specific data 1541 * @tipc_ptr: TIPC specific data 1542 * @atalk_ptr: AppleTalk link 1543 * @ip_ptr: IPv4 specific data 1544 * @dn_ptr: DECnet specific data 1545 * @ip6_ptr: IPv6 specific data 1546 * @ax25_ptr: AX.25 specific data 1547 * @ieee80211_ptr: IEEE 802.11 specific data, assign before registering 1548 * 1549 * @dev_addr: Hw address (before bcast, 1550 * because most packets are unicast) 1551 * 1552 * @_rx: Array of RX queues 1553 * @num_rx_queues: Number of RX queues 1554 * allocated at register_netdev() time 1555 * @real_num_rx_queues: Number of RX queues currently active in device 1556 * 1557 * @rx_handler: handler for received packets 1558 * @rx_handler_data: XXX: need comments on this one 1559 * @ingress_queue: XXX: need comments on this one 1560 * @broadcast: hw bcast address 1561 * 1562 * @rx_cpu_rmap: CPU reverse-mapping for RX completion interrupts, 1563 * indexed by RX queue number. Assigned by driver. 1564 * This must only be set if the ndo_rx_flow_steer 1565 * operation is defined 1566 * @index_hlist: Device index hash chain 1567 * 1568 * @_tx: Array of TX queues 1569 * @num_tx_queues: Number of TX queues allocated at alloc_netdev_mq() time 1570 * @real_num_tx_queues: Number of TX queues currently active in device 1571 * @qdisc: Root qdisc from userspace point of view 1572 * @tx_queue_len: Max frames per queue allowed 1573 * @tx_global_lock: XXX: need comments on this one 1574 * 1575 * @xps_maps: XXX: need comments on this one 1576 * 1577 * @watchdog_timeo: Represents the timeout that is used by 1578 * the watchdog (see dev_watchdog()) 1579 * @watchdog_timer: List of timers 1580 * 1581 * @pcpu_refcnt: Number of references to this device 1582 * @todo_list: Delayed register/unregister 1583 * @link_watch_list: XXX: need comments on this one 1584 * 1585 * @reg_state: Register/unregister state machine 1586 * @dismantle: Device is going to be freed 1587 * @rtnl_link_state: This enum represents the phases of creating 1588 * a new link 1589 * 1590 * @needs_free_netdev: Should unregister perform free_netdev? 1591 * @priv_destructor: Called from unregister 1592 * @npinfo: XXX: need comments on this one 1593 * @nd_net: Network namespace this network device is inside 1594 * 1595 * @ml_priv: Mid-layer private 1596 * @lstats: Loopback statistics 1597 * @tstats: Tunnel statistics 1598 * @dstats: Dummy statistics 1599 * @vstats: Virtual ethernet statistics 1600 * 1601 * @garp_port: GARP 1602 * @mrp_port: MRP 1603 * 1604 * @dev: Class/net/name entry 1605 * @sysfs_groups: Space for optional device, statistics and wireless 1606 * sysfs groups 1607 * 1608 * @sysfs_rx_queue_group: Space for optional per-rx queue attributes 1609 * @rtnl_link_ops: Rtnl_link_ops 1610 * 1611 * @gso_max_size: Maximum size of generic segmentation offload 1612 * @gso_max_segs: Maximum number of segments that can be passed to the 1613 * NIC for GSO 1614 * 1615 * @dcbnl_ops: Data Center Bridging netlink ops 1616 * @num_tc: Number of traffic classes in the net device 1617 * @tc_to_txq: XXX: need comments on this one 1618 * @prio_tc_map: XXX: need comments on this one 1619 * 1620 * @fcoe_ddp_xid: Max exchange id for FCoE LRO by ddp 1621 * 1622 * @priomap: XXX: need comments on this one 1623 * @phydev: Physical device may attach itself 1624 * for hardware timestamping 1625 * 1626 * @qdisc_tx_busylock: lockdep class annotating Qdisc->busylock spinlock 1627 * @qdisc_running_key: lockdep class annotating Qdisc->running seqcount 1628 * 1629 * @proto_down: protocol port state information can be sent to the 1630 * switch driver and used to set the phys state of the 1631 * switch port. 1632 * 1633 * FIXME: cleanup struct net_device such that network protocol info 1634 * moves out. 1635 */ 1636 1637 struct net_device { 1638 char name[IFNAMSIZ]; 1639 struct hlist_node name_hlist; 1640 struct dev_ifalias __rcu *ifalias; 1641 /* 1642 * I/O specific fields 1643 * FIXME: Merge these and struct ifmap into one 1644 */ 1645 unsigned long mem_end; 1646 unsigned long mem_start; 1647 unsigned long base_addr; 1648 int irq; 1649 1650 atomic_t carrier_changes; 1651 1652 /* 1653 * Some hardware also needs these fields (state,dev_list, 1654 * napi_list,unreg_list,close_list) but they are not 1655 * part of the usual set specified in Space.c. 1656 */ 1657 1658 unsigned long state; 1659 1660 struct list_head dev_list; 1661 struct list_head napi_list; 1662 struct list_head unreg_list; 1663 struct list_head close_list; 1664 struct list_head ptype_all; 1665 struct list_head ptype_specific; 1666 1667 struct { 1668 struct list_head upper; 1669 struct list_head lower; 1670 } adj_list; 1671 1672 netdev_features_t features; 1673 netdev_features_t hw_features; 1674 netdev_features_t wanted_features; 1675 netdev_features_t vlan_features; 1676 netdev_features_t hw_enc_features; 1677 netdev_features_t mpls_features; 1678 netdev_features_t gso_partial_features; 1679 1680 int ifindex; 1681 int group; 1682 1683 struct net_device_stats stats; 1684 1685 atomic_long_t rx_dropped; 1686 atomic_long_t tx_dropped; 1687 atomic_long_t rx_nohandler; 1688 1689 #ifdef CONFIG_WIRELESS_EXT 1690 const struct iw_handler_def *wireless_handlers; 1691 struct iw_public_data *wireless_data; 1692 #endif 1693 const struct net_device_ops *netdev_ops; 1694 const struct ethtool_ops *ethtool_ops; 1695 #ifdef CONFIG_NET_SWITCHDEV 1696 const struct switchdev_ops *switchdev_ops; 1697 #endif 1698 #ifdef CONFIG_NET_L3_MASTER_DEV 1699 const struct l3mdev_ops *l3mdev_ops; 1700 #endif 1701 #if IS_ENABLED(CONFIG_IPV6) 1702 const struct ndisc_ops *ndisc_ops; 1703 #endif 1704 1705 #ifdef CONFIG_XFRM 1706 const struct xfrmdev_ops *xfrmdev_ops; 1707 #endif 1708 1709 const struct header_ops *header_ops; 1710 1711 unsigned int flags; 1712 unsigned int priv_flags; 1713 1714 unsigned short gflags; 1715 unsigned short padded; 1716 1717 unsigned char operstate; 1718 unsigned char link_mode; 1719 1720 unsigned char if_port; 1721 unsigned char dma; 1722 1723 unsigned int mtu; 1724 unsigned int min_mtu; 1725 unsigned int max_mtu; 1726 unsigned short type; 1727 unsigned short hard_header_len; 1728 unsigned char min_header_len; 1729 1730 unsigned short needed_headroom; 1731 unsigned short needed_tailroom; 1732 1733 /* Interface address info. */ 1734 unsigned char perm_addr[MAX_ADDR_LEN]; 1735 unsigned char addr_assign_type; 1736 unsigned char addr_len; 1737 unsigned short neigh_priv_len; 1738 unsigned short dev_id; 1739 unsigned short dev_port; 1740 spinlock_t addr_list_lock; 1741 unsigned char name_assign_type; 1742 bool uc_promisc; 1743 struct netdev_hw_addr_list uc; 1744 struct netdev_hw_addr_list mc; 1745 struct netdev_hw_addr_list dev_addrs; 1746 1747 #ifdef CONFIG_SYSFS 1748 struct kset *queues_kset; 1749 #endif 1750 unsigned int promiscuity; 1751 unsigned int allmulti; 1752 1753 1754 /* Protocol-specific pointers */ 1755 1756 #if IS_ENABLED(CONFIG_VLAN_8021Q) 1757 struct vlan_info __rcu *vlan_info; 1758 #endif 1759 #if IS_ENABLED(CONFIG_NET_DSA) 1760 struct dsa_port *dsa_ptr; 1761 #endif 1762 #if IS_ENABLED(CONFIG_TIPC) 1763 struct tipc_bearer __rcu *tipc_ptr; 1764 #endif 1765 void *atalk_ptr; 1766 struct in_device __rcu *ip_ptr; 1767 struct dn_dev __rcu *dn_ptr; 1768 struct inet6_dev __rcu *ip6_ptr; 1769 void *ax25_ptr; 1770 struct wireless_dev *ieee80211_ptr; 1771 struct wpan_dev *ieee802154_ptr; 1772 #if IS_ENABLED(CONFIG_MPLS_ROUTING) 1773 struct mpls_dev __rcu *mpls_ptr; 1774 #endif 1775 1776 /* 1777 * Cache lines mostly used on receive path (including eth_type_trans()) 1778 */ 1779 /* Interface address info used in eth_type_trans() */ 1780 unsigned char *dev_addr; 1781 1782 #ifdef CONFIG_SYSFS 1783 struct netdev_rx_queue *_rx; 1784 1785 unsigned int num_rx_queues; 1786 unsigned int real_num_rx_queues; 1787 #endif 1788 1789 struct bpf_prog __rcu *xdp_prog; 1790 unsigned long gro_flush_timeout; 1791 rx_handler_func_t __rcu *rx_handler; 1792 void __rcu *rx_handler_data; 1793 1794 #ifdef CONFIG_NET_CLS_ACT 1795 struct tcf_proto __rcu *ingress_cl_list; 1796 #endif 1797 struct netdev_queue __rcu *ingress_queue; 1798 #ifdef CONFIG_NETFILTER_INGRESS 1799 struct nf_hook_entries __rcu *nf_hooks_ingress; 1800 #endif 1801 1802 unsigned char broadcast[MAX_ADDR_LEN]; 1803 #ifdef CONFIG_RFS_ACCEL 1804 struct cpu_rmap *rx_cpu_rmap; 1805 #endif 1806 struct hlist_node index_hlist; 1807 1808 /* 1809 * Cache lines mostly used on transmit path 1810 */ 1811 struct netdev_queue *_tx ____cacheline_aligned_in_smp; 1812 unsigned int num_tx_queues; 1813 unsigned int real_num_tx_queues; 1814 struct Qdisc *qdisc; 1815 #ifdef CONFIG_NET_SCHED 1816 DECLARE_HASHTABLE (qdisc_hash, 4); 1817 #endif 1818 unsigned int tx_queue_len; 1819 spinlock_t tx_global_lock; 1820 int watchdog_timeo; 1821 1822 #ifdef CONFIG_XPS 1823 struct xps_dev_maps __rcu *xps_maps; 1824 #endif 1825 #ifdef CONFIG_NET_CLS_ACT 1826 struct tcf_proto __rcu *egress_cl_list; 1827 #endif 1828 1829 /* These may be needed for future network-power-down code. */ 1830 struct timer_list watchdog_timer; 1831 1832 int __percpu *pcpu_refcnt; 1833 struct list_head todo_list; 1834 1835 struct list_head link_watch_list; 1836 1837 enum { NETREG_UNINITIALIZED=0, 1838 NETREG_REGISTERED, /* completed register_netdevice */ 1839 NETREG_UNREGISTERING, /* called unregister_netdevice */ 1840 NETREG_UNREGISTERED, /* completed unregister todo */ 1841 NETREG_RELEASED, /* called free_netdev */ 1842 NETREG_DUMMY, /* dummy device for NAPI poll */ 1843 } reg_state:8; 1844 1845 bool dismantle; 1846 1847 enum { 1848 RTNL_LINK_INITIALIZED, 1849 RTNL_LINK_INITIALIZING, 1850 } rtnl_link_state:16; 1851 1852 bool needs_free_netdev; 1853 void (*priv_destructor)(struct net_device *dev); 1854 1855 #ifdef CONFIG_NETPOLL 1856 struct netpoll_info __rcu *npinfo; 1857 #endif 1858 1859 possible_net_t nd_net; 1860 1861 /* mid-layer private */ 1862 union { 1863 void *ml_priv; 1864 struct pcpu_lstats __percpu *lstats; 1865 struct pcpu_sw_netstats __percpu *tstats; 1866 struct pcpu_dstats __percpu *dstats; 1867 struct pcpu_vstats __percpu *vstats; 1868 }; 1869 1870 #if IS_ENABLED(CONFIG_GARP) 1871 struct garp_port __rcu *garp_port; 1872 #endif 1873 #if IS_ENABLED(CONFIG_MRP) 1874 struct mrp_port __rcu *mrp_port; 1875 #endif 1876 1877 struct device dev; 1878 const struct attribute_group *sysfs_groups[4]; 1879 const struct attribute_group *sysfs_rx_queue_group; 1880 1881 const struct rtnl_link_ops *rtnl_link_ops; 1882 1883 /* for setting kernel sock attribute on TCP connection setup */ 1884 #define GSO_MAX_SIZE 65536 1885 unsigned int gso_max_size; 1886 #define GSO_MAX_SEGS 65535 1887 u16 gso_max_segs; 1888 1889 #ifdef CONFIG_DCB 1890 const struct dcbnl_rtnl_ops *dcbnl_ops; 1891 #endif 1892 u8 num_tc; 1893 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE]; 1894 u8 prio_tc_map[TC_BITMASK + 1]; 1895 1896 #if IS_ENABLED(CONFIG_FCOE) 1897 unsigned int fcoe_ddp_xid; 1898 #endif 1899 #if IS_ENABLED(CONFIG_CGROUP_NET_PRIO) 1900 struct netprio_map __rcu *priomap; 1901 #endif 1902 struct phy_device *phydev; 1903 struct lock_class_key *qdisc_tx_busylock; 1904 struct lock_class_key *qdisc_running_key; 1905 bool proto_down; 1906 }; 1907 #define to_net_dev(d) container_of(d, struct net_device, dev) 1908 1909 static inline bool netif_elide_gro(const struct net_device *dev) 1910 { 1911 if (!(dev->features & NETIF_F_GRO) || dev->xdp_prog) 1912 return true; 1913 return false; 1914 } 1915 1916 #define NETDEV_ALIGN 32 1917 1918 static inline 1919 int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio) 1920 { 1921 return dev->prio_tc_map[prio & TC_BITMASK]; 1922 } 1923 1924 static inline 1925 int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc) 1926 { 1927 if (tc >= dev->num_tc) 1928 return -EINVAL; 1929 1930 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK; 1931 return 0; 1932 } 1933 1934 int netdev_txq_to_tc(struct net_device *dev, unsigned int txq); 1935 void netdev_reset_tc(struct net_device *dev); 1936 int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset); 1937 int netdev_set_num_tc(struct net_device *dev, u8 num_tc); 1938 1939 static inline 1940 int netdev_get_num_tc(struct net_device *dev) 1941 { 1942 return dev->num_tc; 1943 } 1944 1945 static inline 1946 struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev, 1947 unsigned int index) 1948 { 1949 return &dev->_tx[index]; 1950 } 1951 1952 static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev, 1953 const struct sk_buff *skb) 1954 { 1955 return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb)); 1956 } 1957 1958 static inline void netdev_for_each_tx_queue(struct net_device *dev, 1959 void (*f)(struct net_device *, 1960 struct netdev_queue *, 1961 void *), 1962 void *arg) 1963 { 1964 unsigned int i; 1965 1966 for (i = 0; i < dev->num_tx_queues; i++) 1967 f(dev, &dev->_tx[i], arg); 1968 } 1969 1970 #define netdev_lockdep_set_classes(dev) \ 1971 { \ 1972 static struct lock_class_key qdisc_tx_busylock_key; \ 1973 static struct lock_class_key qdisc_running_key; \ 1974 static struct lock_class_key qdisc_xmit_lock_key; \ 1975 static struct lock_class_key dev_addr_list_lock_key; \ 1976 unsigned int i; \ 1977 \ 1978 (dev)->qdisc_tx_busylock = &qdisc_tx_busylock_key; \ 1979 (dev)->qdisc_running_key = &qdisc_running_key; \ 1980 lockdep_set_class(&(dev)->addr_list_lock, \ 1981 &dev_addr_list_lock_key); \ 1982 for (i = 0; i < (dev)->num_tx_queues; i++) \ 1983 lockdep_set_class(&(dev)->_tx[i]._xmit_lock, \ 1984 &qdisc_xmit_lock_key); \ 1985 } 1986 1987 struct netdev_queue *netdev_pick_tx(struct net_device *dev, 1988 struct sk_buff *skb, 1989 void *accel_priv); 1990 1991 /* returns the headroom that the master device needs to take in account 1992 * when forwarding to this dev 1993 */ 1994 static inline unsigned netdev_get_fwd_headroom(struct net_device *dev) 1995 { 1996 return dev->priv_flags & IFF_PHONY_HEADROOM ? 0 : dev->needed_headroom; 1997 } 1998 1999 static inline void netdev_set_rx_headroom(struct net_device *dev, int new_hr) 2000 { 2001 if (dev->netdev_ops->ndo_set_rx_headroom) 2002 dev->netdev_ops->ndo_set_rx_headroom(dev, new_hr); 2003 } 2004 2005 /* set the device rx headroom to the dev's default */ 2006 static inline void netdev_reset_rx_headroom(struct net_device *dev) 2007 { 2008 netdev_set_rx_headroom(dev, -1); 2009 } 2010 2011 /* 2012 * Net namespace inlines 2013 */ 2014 static inline 2015 struct net *dev_net(const struct net_device *dev) 2016 { 2017 return read_pnet(&dev->nd_net); 2018 } 2019 2020 static inline 2021 void dev_net_set(struct net_device *dev, struct net *net) 2022 { 2023 write_pnet(&dev->nd_net, net); 2024 } 2025 2026 /** 2027 * netdev_priv - access network device private data 2028 * @dev: network device 2029 * 2030 * Get network device private data 2031 */ 2032 static inline void *netdev_priv(const struct net_device *dev) 2033 { 2034 return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN); 2035 } 2036 2037 /* Set the sysfs physical device reference for the network logical device 2038 * if set prior to registration will cause a symlink during initialization. 2039 */ 2040 #define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev)) 2041 2042 /* Set the sysfs device type for the network logical device to allow 2043 * fine-grained identification of different network device types. For 2044 * example Ethernet, Wireless LAN, Bluetooth, WiMAX etc. 2045 */ 2046 #define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype)) 2047 2048 /* Default NAPI poll() weight 2049 * Device drivers are strongly advised to not use bigger value 2050 */ 2051 #define NAPI_POLL_WEIGHT 64 2052 2053 /** 2054 * netif_napi_add - initialize a NAPI context 2055 * @dev: network device 2056 * @napi: NAPI context 2057 * @poll: polling function 2058 * @weight: default weight 2059 * 2060 * netif_napi_add() must be used to initialize a NAPI context prior to calling 2061 * *any* of the other NAPI-related functions. 2062 */ 2063 void netif_napi_add(struct net_device *dev, struct napi_struct *napi, 2064 int (*poll)(struct napi_struct *, int), int weight); 2065 2066 /** 2067 * netif_tx_napi_add - initialize a NAPI context 2068 * @dev: network device 2069 * @napi: NAPI context 2070 * @poll: polling function 2071 * @weight: default weight 2072 * 2073 * This variant of netif_napi_add() should be used from drivers using NAPI 2074 * to exclusively poll a TX queue. 2075 * This will avoid we add it into napi_hash[], thus polluting this hash table. 2076 */ 2077 static inline void netif_tx_napi_add(struct net_device *dev, 2078 struct napi_struct *napi, 2079 int (*poll)(struct napi_struct *, int), 2080 int weight) 2081 { 2082 set_bit(NAPI_STATE_NO_BUSY_POLL, &napi->state); 2083 netif_napi_add(dev, napi, poll, weight); 2084 } 2085 2086 /** 2087 * netif_napi_del - remove a NAPI context 2088 * @napi: NAPI context 2089 * 2090 * netif_napi_del() removes a NAPI context from the network device NAPI list 2091 */ 2092 void netif_napi_del(struct napi_struct *napi); 2093 2094 struct napi_gro_cb { 2095 /* Virtual address of skb_shinfo(skb)->frags[0].page + offset. */ 2096 void *frag0; 2097 2098 /* Length of frag0. */ 2099 unsigned int frag0_len; 2100 2101 /* This indicates where we are processing relative to skb->data. */ 2102 int data_offset; 2103 2104 /* This is non-zero if the packet cannot be merged with the new skb. */ 2105 u16 flush; 2106 2107 /* Save the IP ID here and check when we get to the transport layer */ 2108 u16 flush_id; 2109 2110 /* Number of segments aggregated. */ 2111 u16 count; 2112 2113 /* Start offset for remote checksum offload */ 2114 u16 gro_remcsum_start; 2115 2116 /* jiffies when first packet was created/queued */ 2117 unsigned long age; 2118 2119 /* Used in ipv6_gro_receive() and foo-over-udp */ 2120 u16 proto; 2121 2122 /* This is non-zero if the packet may be of the same flow. */ 2123 u8 same_flow:1; 2124 2125 /* Used in tunnel GRO receive */ 2126 u8 encap_mark:1; 2127 2128 /* GRO checksum is valid */ 2129 u8 csum_valid:1; 2130 2131 /* Number of checksums via CHECKSUM_UNNECESSARY */ 2132 u8 csum_cnt:3; 2133 2134 /* Free the skb? */ 2135 u8 free:2; 2136 #define NAPI_GRO_FREE 1 2137 #define NAPI_GRO_FREE_STOLEN_HEAD 2 2138 2139 /* Used in foo-over-udp, set in udp[46]_gro_receive */ 2140 u8 is_ipv6:1; 2141 2142 /* Used in GRE, set in fou/gue_gro_receive */ 2143 u8 is_fou:1; 2144 2145 /* Used to determine if flush_id can be ignored */ 2146 u8 is_atomic:1; 2147 2148 /* Number of gro_receive callbacks this packet already went through */ 2149 u8 recursion_counter:4; 2150 2151 /* 1 bit hole */ 2152 2153 /* used to support CHECKSUM_COMPLETE for tunneling protocols */ 2154 __wsum csum; 2155 2156 /* used in skb_gro_receive() slow path */ 2157 struct sk_buff *last; 2158 }; 2159 2160 #define NAPI_GRO_CB(skb) ((struct napi_gro_cb *)(skb)->cb) 2161 2162 #define GRO_RECURSION_LIMIT 15 2163 static inline int gro_recursion_inc_test(struct sk_buff *skb) 2164 { 2165 return ++NAPI_GRO_CB(skb)->recursion_counter == GRO_RECURSION_LIMIT; 2166 } 2167 2168 typedef struct sk_buff **(*gro_receive_t)(struct sk_buff **, struct sk_buff *); 2169 static inline struct sk_buff **call_gro_receive(gro_receive_t cb, 2170 struct sk_buff **head, 2171 struct sk_buff *skb) 2172 { 2173 if (unlikely(gro_recursion_inc_test(skb))) { 2174 NAPI_GRO_CB(skb)->flush |= 1; 2175 return NULL; 2176 } 2177 2178 return cb(head, skb); 2179 } 2180 2181 typedef struct sk_buff **(*gro_receive_sk_t)(struct sock *, struct sk_buff **, 2182 struct sk_buff *); 2183 static inline struct sk_buff **call_gro_receive_sk(gro_receive_sk_t cb, 2184 struct sock *sk, 2185 struct sk_buff **head, 2186 struct sk_buff *skb) 2187 { 2188 if (unlikely(gro_recursion_inc_test(skb))) { 2189 NAPI_GRO_CB(skb)->flush |= 1; 2190 return NULL; 2191 } 2192 2193 return cb(sk, head, skb); 2194 } 2195 2196 struct packet_type { 2197 __be16 type; /* This is really htons(ether_type). */ 2198 struct net_device *dev; /* NULL is wildcarded here */ 2199 int (*func) (struct sk_buff *, 2200 struct net_device *, 2201 struct packet_type *, 2202 struct net_device *); 2203 bool (*id_match)(struct packet_type *ptype, 2204 struct sock *sk); 2205 void *af_packet_priv; 2206 struct list_head list; 2207 }; 2208 2209 struct offload_callbacks { 2210 struct sk_buff *(*gso_segment)(struct sk_buff *skb, 2211 netdev_features_t features); 2212 struct sk_buff **(*gro_receive)(struct sk_buff **head, 2213 struct sk_buff *skb); 2214 int (*gro_complete)(struct sk_buff *skb, int nhoff); 2215 }; 2216 2217 struct packet_offload { 2218 __be16 type; /* This is really htons(ether_type). */ 2219 u16 priority; 2220 struct offload_callbacks callbacks; 2221 struct list_head list; 2222 }; 2223 2224 /* often modified stats are per-CPU, other are shared (netdev->stats) */ 2225 struct pcpu_sw_netstats { 2226 u64 rx_packets; 2227 u64 rx_bytes; 2228 u64 tx_packets; 2229 u64 tx_bytes; 2230 struct u64_stats_sync syncp; 2231 }; 2232 2233 #define __netdev_alloc_pcpu_stats(type, gfp) \ 2234 ({ \ 2235 typeof(type) __percpu *pcpu_stats = alloc_percpu_gfp(type, gfp);\ 2236 if (pcpu_stats) { \ 2237 int __cpu; \ 2238 for_each_possible_cpu(__cpu) { \ 2239 typeof(type) *stat; \ 2240 stat = per_cpu_ptr(pcpu_stats, __cpu); \ 2241 u64_stats_init(&stat->syncp); \ 2242 } \ 2243 } \ 2244 pcpu_stats; \ 2245 }) 2246 2247 #define netdev_alloc_pcpu_stats(type) \ 2248 __netdev_alloc_pcpu_stats(type, GFP_KERNEL) 2249 2250 enum netdev_lag_tx_type { 2251 NETDEV_LAG_TX_TYPE_UNKNOWN, 2252 NETDEV_LAG_TX_TYPE_RANDOM, 2253 NETDEV_LAG_TX_TYPE_BROADCAST, 2254 NETDEV_LAG_TX_TYPE_ROUNDROBIN, 2255 NETDEV_LAG_TX_TYPE_ACTIVEBACKUP, 2256 NETDEV_LAG_TX_TYPE_HASH, 2257 }; 2258 2259 struct netdev_lag_upper_info { 2260 enum netdev_lag_tx_type tx_type; 2261 }; 2262 2263 struct netdev_lag_lower_state_info { 2264 u8 link_up : 1, 2265 tx_enabled : 1; 2266 }; 2267 2268 #include <linux/notifier.h> 2269 2270 /* netdevice notifier chain. Please remember to update the rtnetlink 2271 * notification exclusion list in rtnetlink_event() when adding new 2272 * types. 2273 */ 2274 #define NETDEV_UP 0x0001 /* For now you can't veto a device up/down */ 2275 #define NETDEV_DOWN 0x0002 2276 #define NETDEV_REBOOT 0x0003 /* Tell a protocol stack a network interface 2277 detected a hardware crash and restarted 2278 - we can use this eg to kick tcp sessions 2279 once done */ 2280 #define NETDEV_CHANGE 0x0004 /* Notify device state change */ 2281 #define NETDEV_REGISTER 0x0005 2282 #define NETDEV_UNREGISTER 0x0006 2283 #define NETDEV_CHANGEMTU 0x0007 /* notify after mtu change happened */ 2284 #define NETDEV_CHANGEADDR 0x0008 2285 #define NETDEV_GOING_DOWN 0x0009 2286 #define NETDEV_CHANGENAME 0x000A 2287 #define NETDEV_FEAT_CHANGE 0x000B 2288 #define NETDEV_BONDING_FAILOVER 0x000C 2289 #define NETDEV_PRE_UP 0x000D 2290 #define NETDEV_PRE_TYPE_CHANGE 0x000E 2291 #define NETDEV_POST_TYPE_CHANGE 0x000F 2292 #define NETDEV_POST_INIT 0x0010 2293 #define NETDEV_UNREGISTER_FINAL 0x0011 2294 #define NETDEV_RELEASE 0x0012 2295 #define NETDEV_NOTIFY_PEERS 0x0013 2296 #define NETDEV_JOIN 0x0014 2297 #define NETDEV_CHANGEUPPER 0x0015 2298 #define NETDEV_RESEND_IGMP 0x0016 2299 #define NETDEV_PRECHANGEMTU 0x0017 /* notify before mtu change happened */ 2300 #define NETDEV_CHANGEINFODATA 0x0018 2301 #define NETDEV_BONDING_INFO 0x0019 2302 #define NETDEV_PRECHANGEUPPER 0x001A 2303 #define NETDEV_CHANGELOWERSTATE 0x001B 2304 #define NETDEV_UDP_TUNNEL_PUSH_INFO 0x001C 2305 #define NETDEV_UDP_TUNNEL_DROP_INFO 0x001D 2306 #define NETDEV_CHANGE_TX_QUEUE_LEN 0x001E 2307 2308 int register_netdevice_notifier(struct notifier_block *nb); 2309 int unregister_netdevice_notifier(struct notifier_block *nb); 2310 2311 struct netdev_notifier_info { 2312 struct net_device *dev; 2313 }; 2314 2315 struct netdev_notifier_change_info { 2316 struct netdev_notifier_info info; /* must be first */ 2317 unsigned int flags_changed; 2318 }; 2319 2320 struct netdev_notifier_changeupper_info { 2321 struct netdev_notifier_info info; /* must be first */ 2322 struct net_device *upper_dev; /* new upper dev */ 2323 bool master; /* is upper dev master */ 2324 bool linking; /* is the notification for link or unlink */ 2325 void *upper_info; /* upper dev info */ 2326 }; 2327 2328 struct netdev_notifier_changelowerstate_info { 2329 struct netdev_notifier_info info; /* must be first */ 2330 void *lower_state_info; /* is lower dev state */ 2331 }; 2332 2333 static inline void netdev_notifier_info_init(struct netdev_notifier_info *info, 2334 struct net_device *dev) 2335 { 2336 info->dev = dev; 2337 } 2338 2339 static inline struct net_device * 2340 netdev_notifier_info_to_dev(const struct netdev_notifier_info *info) 2341 { 2342 return info->dev; 2343 } 2344 2345 int call_netdevice_notifiers(unsigned long val, struct net_device *dev); 2346 2347 2348 extern rwlock_t dev_base_lock; /* Device list lock */ 2349 2350 #define for_each_netdev(net, d) \ 2351 list_for_each_entry(d, &(net)->dev_base_head, dev_list) 2352 #define for_each_netdev_reverse(net, d) \ 2353 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list) 2354 #define for_each_netdev_rcu(net, d) \ 2355 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list) 2356 #define for_each_netdev_safe(net, d, n) \ 2357 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list) 2358 #define for_each_netdev_continue(net, d) \ 2359 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list) 2360 #define for_each_netdev_continue_rcu(net, d) \ 2361 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list) 2362 #define for_each_netdev_in_bond_rcu(bond, slave) \ 2363 for_each_netdev_rcu(&init_net, slave) \ 2364 if (netdev_master_upper_dev_get_rcu(slave) == (bond)) 2365 #define net_device_entry(lh) list_entry(lh, struct net_device, dev_list) 2366 2367 static inline struct net_device *next_net_device(struct net_device *dev) 2368 { 2369 struct list_head *lh; 2370 struct net *net; 2371 2372 net = dev_net(dev); 2373 lh = dev->dev_list.next; 2374 return lh == &net->dev_base_head ? NULL : net_device_entry(lh); 2375 } 2376 2377 static inline struct net_device *next_net_device_rcu(struct net_device *dev) 2378 { 2379 struct list_head *lh; 2380 struct net *net; 2381 2382 net = dev_net(dev); 2383 lh = rcu_dereference(list_next_rcu(&dev->dev_list)); 2384 return lh == &net->dev_base_head ? NULL : net_device_entry(lh); 2385 } 2386 2387 static inline struct net_device *first_net_device(struct net *net) 2388 { 2389 return list_empty(&net->dev_base_head) ? NULL : 2390 net_device_entry(net->dev_base_head.next); 2391 } 2392 2393 static inline struct net_device *first_net_device_rcu(struct net *net) 2394 { 2395 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head)); 2396 2397 return lh == &net->dev_base_head ? NULL : net_device_entry(lh); 2398 } 2399 2400 int netdev_boot_setup_check(struct net_device *dev); 2401 unsigned long netdev_boot_base(const char *prefix, int unit); 2402 struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type, 2403 const char *hwaddr); 2404 struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type); 2405 struct net_device *__dev_getfirstbyhwtype(struct net *net, unsigned short type); 2406 void dev_add_pack(struct packet_type *pt); 2407 void dev_remove_pack(struct packet_type *pt); 2408 void __dev_remove_pack(struct packet_type *pt); 2409 void dev_add_offload(struct packet_offload *po); 2410 void dev_remove_offload(struct packet_offload *po); 2411 2412 int dev_get_iflink(const struct net_device *dev); 2413 int dev_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb); 2414 struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags, 2415 unsigned short mask); 2416 struct net_device *dev_get_by_name(struct net *net, const char *name); 2417 struct net_device *dev_get_by_name_rcu(struct net *net, const char *name); 2418 struct net_device *__dev_get_by_name(struct net *net, const char *name); 2419 int dev_alloc_name(struct net_device *dev, const char *name); 2420 int dev_open(struct net_device *dev); 2421 void dev_close(struct net_device *dev); 2422 void dev_close_many(struct list_head *head, bool unlink); 2423 void dev_disable_lro(struct net_device *dev); 2424 int dev_loopback_xmit(struct net *net, struct sock *sk, struct sk_buff *newskb); 2425 int dev_queue_xmit(struct sk_buff *skb); 2426 int dev_queue_xmit_accel(struct sk_buff *skb, void *accel_priv); 2427 int register_netdevice(struct net_device *dev); 2428 void unregister_netdevice_queue(struct net_device *dev, struct list_head *head); 2429 void unregister_netdevice_many(struct list_head *head); 2430 static inline void unregister_netdevice(struct net_device *dev) 2431 { 2432 unregister_netdevice_queue(dev, NULL); 2433 } 2434 2435 int netdev_refcnt_read(const struct net_device *dev); 2436 void free_netdev(struct net_device *dev); 2437 void netdev_freemem(struct net_device *dev); 2438 void synchronize_net(void); 2439 int init_dummy_netdev(struct net_device *dev); 2440 2441 DECLARE_PER_CPU(int, xmit_recursion); 2442 #define XMIT_RECURSION_LIMIT 10 2443 2444 static inline int dev_recursion_level(void) 2445 { 2446 return this_cpu_read(xmit_recursion); 2447 } 2448 2449 struct net_device *dev_get_by_index(struct net *net, int ifindex); 2450 struct net_device *__dev_get_by_index(struct net *net, int ifindex); 2451 struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex); 2452 struct net_device *dev_get_by_napi_id(unsigned int napi_id); 2453 int netdev_get_name(struct net *net, char *name, int ifindex); 2454 int dev_restart(struct net_device *dev); 2455 int skb_gro_receive(struct sk_buff **head, struct sk_buff *skb); 2456 2457 static inline unsigned int skb_gro_offset(const struct sk_buff *skb) 2458 { 2459 return NAPI_GRO_CB(skb)->data_offset; 2460 } 2461 2462 static inline unsigned int skb_gro_len(const struct sk_buff *skb) 2463 { 2464 return skb->len - NAPI_GRO_CB(skb)->data_offset; 2465 } 2466 2467 static inline void skb_gro_pull(struct sk_buff *skb, unsigned int len) 2468 { 2469 NAPI_GRO_CB(skb)->data_offset += len; 2470 } 2471 2472 static inline void *skb_gro_header_fast(struct sk_buff *skb, 2473 unsigned int offset) 2474 { 2475 return NAPI_GRO_CB(skb)->frag0 + offset; 2476 } 2477 2478 static inline int skb_gro_header_hard(struct sk_buff *skb, unsigned int hlen) 2479 { 2480 return NAPI_GRO_CB(skb)->frag0_len < hlen; 2481 } 2482 2483 static inline void skb_gro_frag0_invalidate(struct sk_buff *skb) 2484 { 2485 NAPI_GRO_CB(skb)->frag0 = NULL; 2486 NAPI_GRO_CB(skb)->frag0_len = 0; 2487 } 2488 2489 static inline void *skb_gro_header_slow(struct sk_buff *skb, unsigned int hlen, 2490 unsigned int offset) 2491 { 2492 if (!pskb_may_pull(skb, hlen)) 2493 return NULL; 2494 2495 skb_gro_frag0_invalidate(skb); 2496 return skb->data + offset; 2497 } 2498 2499 static inline void *skb_gro_network_header(struct sk_buff *skb) 2500 { 2501 return (NAPI_GRO_CB(skb)->frag0 ?: skb->data) + 2502 skb_network_offset(skb); 2503 } 2504 2505 static inline void skb_gro_postpull_rcsum(struct sk_buff *skb, 2506 const void *start, unsigned int len) 2507 { 2508 if (NAPI_GRO_CB(skb)->csum_valid) 2509 NAPI_GRO_CB(skb)->csum = csum_sub(NAPI_GRO_CB(skb)->csum, 2510 csum_partial(start, len, 0)); 2511 } 2512 2513 /* GRO checksum functions. These are logical equivalents of the normal 2514 * checksum functions (in skbuff.h) except that they operate on the GRO 2515 * offsets and fields in sk_buff. 2516 */ 2517 2518 __sum16 __skb_gro_checksum_complete(struct sk_buff *skb); 2519 2520 static inline bool skb_at_gro_remcsum_start(struct sk_buff *skb) 2521 { 2522 return (NAPI_GRO_CB(skb)->gro_remcsum_start == skb_gro_offset(skb)); 2523 } 2524 2525 static inline bool __skb_gro_checksum_validate_needed(struct sk_buff *skb, 2526 bool zero_okay, 2527 __sum16 check) 2528 { 2529 return ((skb->ip_summed != CHECKSUM_PARTIAL || 2530 skb_checksum_start_offset(skb) < 2531 skb_gro_offset(skb)) && 2532 !skb_at_gro_remcsum_start(skb) && 2533 NAPI_GRO_CB(skb)->csum_cnt == 0 && 2534 (!zero_okay || check)); 2535 } 2536 2537 static inline __sum16 __skb_gro_checksum_validate_complete(struct sk_buff *skb, 2538 __wsum psum) 2539 { 2540 if (NAPI_GRO_CB(skb)->csum_valid && 2541 !csum_fold(csum_add(psum, NAPI_GRO_CB(skb)->csum))) 2542 return 0; 2543 2544 NAPI_GRO_CB(skb)->csum = psum; 2545 2546 return __skb_gro_checksum_complete(skb); 2547 } 2548 2549 static inline void skb_gro_incr_csum_unnecessary(struct sk_buff *skb) 2550 { 2551 if (NAPI_GRO_CB(skb)->csum_cnt > 0) { 2552 /* Consume a checksum from CHECKSUM_UNNECESSARY */ 2553 NAPI_GRO_CB(skb)->csum_cnt--; 2554 } else { 2555 /* Update skb for CHECKSUM_UNNECESSARY and csum_level when we 2556 * verified a new top level checksum or an encapsulated one 2557 * during GRO. This saves work if we fallback to normal path. 2558 */ 2559 __skb_incr_checksum_unnecessary(skb); 2560 } 2561 } 2562 2563 #define __skb_gro_checksum_validate(skb, proto, zero_okay, check, \ 2564 compute_pseudo) \ 2565 ({ \ 2566 __sum16 __ret = 0; \ 2567 if (__skb_gro_checksum_validate_needed(skb, zero_okay, check)) \ 2568 __ret = __skb_gro_checksum_validate_complete(skb, \ 2569 compute_pseudo(skb, proto)); \ 2570 if (!__ret) \ 2571 skb_gro_incr_csum_unnecessary(skb); \ 2572 __ret; \ 2573 }) 2574 2575 #define skb_gro_checksum_validate(skb, proto, compute_pseudo) \ 2576 __skb_gro_checksum_validate(skb, proto, false, 0, compute_pseudo) 2577 2578 #define skb_gro_checksum_validate_zero_check(skb, proto, check, \ 2579 compute_pseudo) \ 2580 __skb_gro_checksum_validate(skb, proto, true, check, compute_pseudo) 2581 2582 #define skb_gro_checksum_simple_validate(skb) \ 2583 __skb_gro_checksum_validate(skb, 0, false, 0, null_compute_pseudo) 2584 2585 static inline bool __skb_gro_checksum_convert_check(struct sk_buff *skb) 2586 { 2587 return (NAPI_GRO_CB(skb)->csum_cnt == 0 && 2588 !NAPI_GRO_CB(skb)->csum_valid); 2589 } 2590 2591 static inline void __skb_gro_checksum_convert(struct sk_buff *skb, 2592 __sum16 check, __wsum pseudo) 2593 { 2594 NAPI_GRO_CB(skb)->csum = ~pseudo; 2595 NAPI_GRO_CB(skb)->csum_valid = 1; 2596 } 2597 2598 #define skb_gro_checksum_try_convert(skb, proto, check, compute_pseudo) \ 2599 do { \ 2600 if (__skb_gro_checksum_convert_check(skb)) \ 2601 __skb_gro_checksum_convert(skb, check, \ 2602 compute_pseudo(skb, proto)); \ 2603 } while (0) 2604 2605 struct gro_remcsum { 2606 int offset; 2607 __wsum delta; 2608 }; 2609 2610 static inline void skb_gro_remcsum_init(struct gro_remcsum *grc) 2611 { 2612 grc->offset = 0; 2613 grc->delta = 0; 2614 } 2615 2616 static inline void *skb_gro_remcsum_process(struct sk_buff *skb, void *ptr, 2617 unsigned int off, size_t hdrlen, 2618 int start, int offset, 2619 struct gro_remcsum *grc, 2620 bool nopartial) 2621 { 2622 __wsum delta; 2623 size_t plen = hdrlen + max_t(size_t, offset + sizeof(u16), start); 2624 2625 BUG_ON(!NAPI_GRO_CB(skb)->csum_valid); 2626 2627 if (!nopartial) { 2628 NAPI_GRO_CB(skb)->gro_remcsum_start = off + hdrlen + start; 2629 return ptr; 2630 } 2631 2632 ptr = skb_gro_header_fast(skb, off); 2633 if (skb_gro_header_hard(skb, off + plen)) { 2634 ptr = skb_gro_header_slow(skb, off + plen, off); 2635 if (!ptr) 2636 return NULL; 2637 } 2638 2639 delta = remcsum_adjust(ptr + hdrlen, NAPI_GRO_CB(skb)->csum, 2640 start, offset); 2641 2642 /* Adjust skb->csum since we changed the packet */ 2643 NAPI_GRO_CB(skb)->csum = csum_add(NAPI_GRO_CB(skb)->csum, delta); 2644 2645 grc->offset = off + hdrlen + offset; 2646 grc->delta = delta; 2647 2648 return ptr; 2649 } 2650 2651 static inline void skb_gro_remcsum_cleanup(struct sk_buff *skb, 2652 struct gro_remcsum *grc) 2653 { 2654 void *ptr; 2655 size_t plen = grc->offset + sizeof(u16); 2656 2657 if (!grc->delta) 2658 return; 2659 2660 ptr = skb_gro_header_fast(skb, grc->offset); 2661 if (skb_gro_header_hard(skb, grc->offset + sizeof(u16))) { 2662 ptr = skb_gro_header_slow(skb, plen, grc->offset); 2663 if (!ptr) 2664 return; 2665 } 2666 2667 remcsum_unadjust((__sum16 *)ptr, grc->delta); 2668 } 2669 2670 #ifdef CONFIG_XFRM_OFFLOAD 2671 static inline void skb_gro_flush_final(struct sk_buff *skb, struct sk_buff **pp, int flush) 2672 { 2673 if (PTR_ERR(pp) != -EINPROGRESS) 2674 NAPI_GRO_CB(skb)->flush |= flush; 2675 } 2676 #else 2677 static inline void skb_gro_flush_final(struct sk_buff *skb, struct sk_buff **pp, int flush) 2678 { 2679 NAPI_GRO_CB(skb)->flush |= flush; 2680 } 2681 #endif 2682 2683 static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev, 2684 unsigned short type, 2685 const void *daddr, const void *saddr, 2686 unsigned int len) 2687 { 2688 if (!dev->header_ops || !dev->header_ops->create) 2689 return 0; 2690 2691 return dev->header_ops->create(skb, dev, type, daddr, saddr, len); 2692 } 2693 2694 static inline int dev_parse_header(const struct sk_buff *skb, 2695 unsigned char *haddr) 2696 { 2697 const struct net_device *dev = skb->dev; 2698 2699 if (!dev->header_ops || !dev->header_ops->parse) 2700 return 0; 2701 return dev->header_ops->parse(skb, haddr); 2702 } 2703 2704 /* ll_header must have at least hard_header_len allocated */ 2705 static inline bool dev_validate_header(const struct net_device *dev, 2706 char *ll_header, int len) 2707 { 2708 if (likely(len >= dev->hard_header_len)) 2709 return true; 2710 if (len < dev->min_header_len) 2711 return false; 2712 2713 if (capable(CAP_SYS_RAWIO)) { 2714 memset(ll_header + len, 0, dev->hard_header_len - len); 2715 return true; 2716 } 2717 2718 if (dev->header_ops && dev->header_ops->validate) 2719 return dev->header_ops->validate(ll_header, len); 2720 2721 return false; 2722 } 2723 2724 typedef int gifconf_func_t(struct net_device * dev, char __user * bufptr, int len); 2725 int register_gifconf(unsigned int family, gifconf_func_t *gifconf); 2726 static inline int unregister_gifconf(unsigned int family) 2727 { 2728 return register_gifconf(family, NULL); 2729 } 2730 2731 #ifdef CONFIG_NET_FLOW_LIMIT 2732 #define FLOW_LIMIT_HISTORY (1 << 7) /* must be ^2 and !overflow buckets */ 2733 struct sd_flow_limit { 2734 u64 count; 2735 unsigned int num_buckets; 2736 unsigned int history_head; 2737 u16 history[FLOW_LIMIT_HISTORY]; 2738 u8 buckets[]; 2739 }; 2740 2741 extern int netdev_flow_limit_table_len; 2742 #endif /* CONFIG_NET_FLOW_LIMIT */ 2743 2744 /* 2745 * Incoming packets are placed on per-CPU queues 2746 */ 2747 struct softnet_data { 2748 struct list_head poll_list; 2749 struct sk_buff_head process_queue; 2750 2751 /* stats */ 2752 unsigned int processed; 2753 unsigned int time_squeeze; 2754 unsigned int received_rps; 2755 #ifdef CONFIG_RPS 2756 struct softnet_data *rps_ipi_list; 2757 #endif 2758 #ifdef CONFIG_NET_FLOW_LIMIT 2759 struct sd_flow_limit __rcu *flow_limit; 2760 #endif 2761 struct Qdisc *output_queue; 2762 struct Qdisc **output_queue_tailp; 2763 struct sk_buff *completion_queue; 2764 2765 #ifdef CONFIG_RPS 2766 /* input_queue_head should be written by cpu owning this struct, 2767 * and only read by other cpus. Worth using a cache line. 2768 */ 2769 unsigned int input_queue_head ____cacheline_aligned_in_smp; 2770 2771 /* Elements below can be accessed between CPUs for RPS/RFS */ 2772 call_single_data_t csd ____cacheline_aligned_in_smp; 2773 struct softnet_data *rps_ipi_next; 2774 unsigned int cpu; 2775 unsigned int input_queue_tail; 2776 #endif 2777 unsigned int dropped; 2778 struct sk_buff_head input_pkt_queue; 2779 struct napi_struct backlog; 2780 2781 }; 2782 2783 static inline void input_queue_head_incr(struct softnet_data *sd) 2784 { 2785 #ifdef CONFIG_RPS 2786 sd->input_queue_head++; 2787 #endif 2788 } 2789 2790 static inline void input_queue_tail_incr_save(struct softnet_data *sd, 2791 unsigned int *qtail) 2792 { 2793 #ifdef CONFIG_RPS 2794 *qtail = ++sd->input_queue_tail; 2795 #endif 2796 } 2797 2798 DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data); 2799 2800 void __netif_schedule(struct Qdisc *q); 2801 void netif_schedule_queue(struct netdev_queue *txq); 2802 2803 static inline void netif_tx_schedule_all(struct net_device *dev) 2804 { 2805 unsigned int i; 2806 2807 for (i = 0; i < dev->num_tx_queues; i++) 2808 netif_schedule_queue(netdev_get_tx_queue(dev, i)); 2809 } 2810 2811 static __always_inline void netif_tx_start_queue(struct netdev_queue *dev_queue) 2812 { 2813 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state); 2814 } 2815 2816 /** 2817 * netif_start_queue - allow transmit 2818 * @dev: network device 2819 * 2820 * Allow upper layers to call the device hard_start_xmit routine. 2821 */ 2822 static inline void netif_start_queue(struct net_device *dev) 2823 { 2824 netif_tx_start_queue(netdev_get_tx_queue(dev, 0)); 2825 } 2826 2827 static inline void netif_tx_start_all_queues(struct net_device *dev) 2828 { 2829 unsigned int i; 2830 2831 for (i = 0; i < dev->num_tx_queues; i++) { 2832 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 2833 netif_tx_start_queue(txq); 2834 } 2835 } 2836 2837 void netif_tx_wake_queue(struct netdev_queue *dev_queue); 2838 2839 /** 2840 * netif_wake_queue - restart transmit 2841 * @dev: network device 2842 * 2843 * Allow upper layers to call the device hard_start_xmit routine. 2844 * Used for flow control when transmit resources are available. 2845 */ 2846 static inline void netif_wake_queue(struct net_device *dev) 2847 { 2848 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0)); 2849 } 2850 2851 static inline void netif_tx_wake_all_queues(struct net_device *dev) 2852 { 2853 unsigned int i; 2854 2855 for (i = 0; i < dev->num_tx_queues; i++) { 2856 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 2857 netif_tx_wake_queue(txq); 2858 } 2859 } 2860 2861 static __always_inline void netif_tx_stop_queue(struct netdev_queue *dev_queue) 2862 { 2863 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state); 2864 } 2865 2866 /** 2867 * netif_stop_queue - stop transmitted packets 2868 * @dev: network device 2869 * 2870 * Stop upper layers calling the device hard_start_xmit routine. 2871 * Used for flow control when transmit resources are unavailable. 2872 */ 2873 static inline void netif_stop_queue(struct net_device *dev) 2874 { 2875 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0)); 2876 } 2877 2878 void netif_tx_stop_all_queues(struct net_device *dev); 2879 2880 static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue) 2881 { 2882 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state); 2883 } 2884 2885 /** 2886 * netif_queue_stopped - test if transmit queue is flowblocked 2887 * @dev: network device 2888 * 2889 * Test if transmit queue on device is currently unable to send. 2890 */ 2891 static inline bool netif_queue_stopped(const struct net_device *dev) 2892 { 2893 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0)); 2894 } 2895 2896 static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue) 2897 { 2898 return dev_queue->state & QUEUE_STATE_ANY_XOFF; 2899 } 2900 2901 static inline bool 2902 netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue) 2903 { 2904 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN; 2905 } 2906 2907 static inline bool 2908 netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue) 2909 { 2910 return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN; 2911 } 2912 2913 /** 2914 * netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write 2915 * @dev_queue: pointer to transmit queue 2916 * 2917 * BQL enabled drivers might use this helper in their ndo_start_xmit(), 2918 * to give appropriate hint to the CPU. 2919 */ 2920 static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue) 2921 { 2922 #ifdef CONFIG_BQL 2923 prefetchw(&dev_queue->dql.num_queued); 2924 #endif 2925 } 2926 2927 /** 2928 * netdev_txq_bql_complete_prefetchw - prefetch bql data for write 2929 * @dev_queue: pointer to transmit queue 2930 * 2931 * BQL enabled drivers might use this helper in their TX completion path, 2932 * to give appropriate hint to the CPU. 2933 */ 2934 static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue) 2935 { 2936 #ifdef CONFIG_BQL 2937 prefetchw(&dev_queue->dql.limit); 2938 #endif 2939 } 2940 2941 static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue, 2942 unsigned int bytes) 2943 { 2944 #ifdef CONFIG_BQL 2945 dql_queued(&dev_queue->dql, bytes); 2946 2947 if (likely(dql_avail(&dev_queue->dql) >= 0)) 2948 return; 2949 2950 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state); 2951 2952 /* 2953 * The XOFF flag must be set before checking the dql_avail below, 2954 * because in netdev_tx_completed_queue we update the dql_completed 2955 * before checking the XOFF flag. 2956 */ 2957 smp_mb(); 2958 2959 /* check again in case another CPU has just made room avail */ 2960 if (unlikely(dql_avail(&dev_queue->dql) >= 0)) 2961 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state); 2962 #endif 2963 } 2964 2965 /** 2966 * netdev_sent_queue - report the number of bytes queued to hardware 2967 * @dev: network device 2968 * @bytes: number of bytes queued to the hardware device queue 2969 * 2970 * Report the number of bytes queued for sending/completion to the network 2971 * device hardware queue. @bytes should be a good approximation and should 2972 * exactly match netdev_completed_queue() @bytes 2973 */ 2974 static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes) 2975 { 2976 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes); 2977 } 2978 2979 static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue, 2980 unsigned int pkts, unsigned int bytes) 2981 { 2982 #ifdef CONFIG_BQL 2983 if (unlikely(!bytes)) 2984 return; 2985 2986 dql_completed(&dev_queue->dql, bytes); 2987 2988 /* 2989 * Without the memory barrier there is a small possiblity that 2990 * netdev_tx_sent_queue will miss the update and cause the queue to 2991 * be stopped forever 2992 */ 2993 smp_mb(); 2994 2995 if (dql_avail(&dev_queue->dql) < 0) 2996 return; 2997 2998 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state)) 2999 netif_schedule_queue(dev_queue); 3000 #endif 3001 } 3002 3003 /** 3004 * netdev_completed_queue - report bytes and packets completed by device 3005 * @dev: network device 3006 * @pkts: actual number of packets sent over the medium 3007 * @bytes: actual number of bytes sent over the medium 3008 * 3009 * Report the number of bytes and packets transmitted by the network device 3010 * hardware queue over the physical medium, @bytes must exactly match the 3011 * @bytes amount passed to netdev_sent_queue() 3012 */ 3013 static inline void netdev_completed_queue(struct net_device *dev, 3014 unsigned int pkts, unsigned int bytes) 3015 { 3016 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes); 3017 } 3018 3019 static inline void netdev_tx_reset_queue(struct netdev_queue *q) 3020 { 3021 #ifdef CONFIG_BQL 3022 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state); 3023 dql_reset(&q->dql); 3024 #endif 3025 } 3026 3027 /** 3028 * netdev_reset_queue - reset the packets and bytes count of a network device 3029 * @dev_queue: network device 3030 * 3031 * Reset the bytes and packet count of a network device and clear the 3032 * software flow control OFF bit for this network device 3033 */ 3034 static inline void netdev_reset_queue(struct net_device *dev_queue) 3035 { 3036 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0)); 3037 } 3038 3039 /** 3040 * netdev_cap_txqueue - check if selected tx queue exceeds device queues 3041 * @dev: network device 3042 * @queue_index: given tx queue index 3043 * 3044 * Returns 0 if given tx queue index >= number of device tx queues, 3045 * otherwise returns the originally passed tx queue index. 3046 */ 3047 static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index) 3048 { 3049 if (unlikely(queue_index >= dev->real_num_tx_queues)) { 3050 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n", 3051 dev->name, queue_index, 3052 dev->real_num_tx_queues); 3053 return 0; 3054 } 3055 3056 return queue_index; 3057 } 3058 3059 /** 3060 * netif_running - test if up 3061 * @dev: network device 3062 * 3063 * Test if the device has been brought up. 3064 */ 3065 static inline bool netif_running(const struct net_device *dev) 3066 { 3067 return test_bit(__LINK_STATE_START, &dev->state); 3068 } 3069 3070 /* 3071 * Routines to manage the subqueues on a device. We only need start, 3072 * stop, and a check if it's stopped. All other device management is 3073 * done at the overall netdevice level. 3074 * Also test the device if we're multiqueue. 3075 */ 3076 3077 /** 3078 * netif_start_subqueue - allow sending packets on subqueue 3079 * @dev: network device 3080 * @queue_index: sub queue index 3081 * 3082 * Start individual transmit queue of a device with multiple transmit queues. 3083 */ 3084 static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index) 3085 { 3086 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 3087 3088 netif_tx_start_queue(txq); 3089 } 3090 3091 /** 3092 * netif_stop_subqueue - stop sending packets on subqueue 3093 * @dev: network device 3094 * @queue_index: sub queue index 3095 * 3096 * Stop individual transmit queue of a device with multiple transmit queues. 3097 */ 3098 static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index) 3099 { 3100 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 3101 netif_tx_stop_queue(txq); 3102 } 3103 3104 /** 3105 * netif_subqueue_stopped - test status of subqueue 3106 * @dev: network device 3107 * @queue_index: sub queue index 3108 * 3109 * Check individual transmit queue of a device with multiple transmit queues. 3110 */ 3111 static inline bool __netif_subqueue_stopped(const struct net_device *dev, 3112 u16 queue_index) 3113 { 3114 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 3115 3116 return netif_tx_queue_stopped(txq); 3117 } 3118 3119 static inline bool netif_subqueue_stopped(const struct net_device *dev, 3120 struct sk_buff *skb) 3121 { 3122 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb)); 3123 } 3124 3125 /** 3126 * netif_wake_subqueue - allow sending packets on subqueue 3127 * @dev: network device 3128 * @queue_index: sub queue index 3129 * 3130 * Resume individual transmit queue of a device with multiple transmit queues. 3131 */ 3132 static inline void netif_wake_subqueue(struct net_device *dev, u16 queue_index) 3133 { 3134 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 3135 3136 netif_tx_wake_queue(txq); 3137 } 3138 3139 #ifdef CONFIG_XPS 3140 int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask, 3141 u16 index); 3142 #else 3143 static inline int netif_set_xps_queue(struct net_device *dev, 3144 const struct cpumask *mask, 3145 u16 index) 3146 { 3147 return 0; 3148 } 3149 #endif 3150 3151 u16 __skb_tx_hash(const struct net_device *dev, struct sk_buff *skb, 3152 unsigned int num_tx_queues); 3153 3154 /* 3155 * Returns a Tx hash for the given packet when dev->real_num_tx_queues is used 3156 * as a distribution range limit for the returned value. 3157 */ 3158 static inline u16 skb_tx_hash(const struct net_device *dev, 3159 struct sk_buff *skb) 3160 { 3161 return __skb_tx_hash(dev, skb, dev->real_num_tx_queues); 3162 } 3163 3164 /** 3165 * netif_is_multiqueue - test if device has multiple transmit queues 3166 * @dev: network device 3167 * 3168 * Check if device has multiple transmit queues 3169 */ 3170 static inline bool netif_is_multiqueue(const struct net_device *dev) 3171 { 3172 return dev->num_tx_queues > 1; 3173 } 3174 3175 int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq); 3176 3177 #ifdef CONFIG_SYSFS 3178 int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq); 3179 #else 3180 static inline int netif_set_real_num_rx_queues(struct net_device *dev, 3181 unsigned int rxq) 3182 { 3183 return 0; 3184 } 3185 #endif 3186 3187 #ifdef CONFIG_SYSFS 3188 static inline unsigned int get_netdev_rx_queue_index( 3189 struct netdev_rx_queue *queue) 3190 { 3191 struct net_device *dev = queue->dev; 3192 int index = queue - dev->_rx; 3193 3194 BUG_ON(index >= dev->num_rx_queues); 3195 return index; 3196 } 3197 #endif 3198 3199 #define DEFAULT_MAX_NUM_RSS_QUEUES (8) 3200 int netif_get_num_default_rss_queues(void); 3201 3202 enum skb_free_reason { 3203 SKB_REASON_CONSUMED, 3204 SKB_REASON_DROPPED, 3205 }; 3206 3207 void __dev_kfree_skb_irq(struct sk_buff *skb, enum skb_free_reason reason); 3208 void __dev_kfree_skb_any(struct sk_buff *skb, enum skb_free_reason reason); 3209 3210 /* 3211 * It is not allowed to call kfree_skb() or consume_skb() from hardware 3212 * interrupt context or with hardware interrupts being disabled. 3213 * (in_irq() || irqs_disabled()) 3214 * 3215 * We provide four helpers that can be used in following contexts : 3216 * 3217 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context, 3218 * replacing kfree_skb(skb) 3219 * 3220 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context. 3221 * Typically used in place of consume_skb(skb) in TX completion path 3222 * 3223 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context, 3224 * replacing kfree_skb(skb) 3225 * 3226 * dev_consume_skb_any(skb) when caller doesn't know its current irq context, 3227 * and consumed a packet. Used in place of consume_skb(skb) 3228 */ 3229 static inline void dev_kfree_skb_irq(struct sk_buff *skb) 3230 { 3231 __dev_kfree_skb_irq(skb, SKB_REASON_DROPPED); 3232 } 3233 3234 static inline void dev_consume_skb_irq(struct sk_buff *skb) 3235 { 3236 __dev_kfree_skb_irq(skb, SKB_REASON_CONSUMED); 3237 } 3238 3239 static inline void dev_kfree_skb_any(struct sk_buff *skb) 3240 { 3241 __dev_kfree_skb_any(skb, SKB_REASON_DROPPED); 3242 } 3243 3244 static inline void dev_consume_skb_any(struct sk_buff *skb) 3245 { 3246 __dev_kfree_skb_any(skb, SKB_REASON_CONSUMED); 3247 } 3248 3249 void generic_xdp_tx(struct sk_buff *skb, struct bpf_prog *xdp_prog); 3250 int do_xdp_generic(struct bpf_prog *xdp_prog, struct sk_buff *skb); 3251 int netif_rx(struct sk_buff *skb); 3252 int netif_rx_ni(struct sk_buff *skb); 3253 int netif_receive_skb(struct sk_buff *skb); 3254 gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb); 3255 void napi_gro_flush(struct napi_struct *napi, bool flush_old); 3256 struct sk_buff *napi_get_frags(struct napi_struct *napi); 3257 gro_result_t napi_gro_frags(struct napi_struct *napi); 3258 struct packet_offload *gro_find_receive_by_type(__be16 type); 3259 struct packet_offload *gro_find_complete_by_type(__be16 type); 3260 3261 static inline void napi_free_frags(struct napi_struct *napi) 3262 { 3263 kfree_skb(napi->skb); 3264 napi->skb = NULL; 3265 } 3266 3267 bool netdev_is_rx_handler_busy(struct net_device *dev); 3268 int netdev_rx_handler_register(struct net_device *dev, 3269 rx_handler_func_t *rx_handler, 3270 void *rx_handler_data); 3271 void netdev_rx_handler_unregister(struct net_device *dev); 3272 3273 bool dev_valid_name(const char *name); 3274 int dev_ioctl(struct net *net, unsigned int cmd, void __user *); 3275 int dev_ethtool(struct net *net, struct ifreq *); 3276 unsigned int dev_get_flags(const struct net_device *); 3277 int __dev_change_flags(struct net_device *, unsigned int flags); 3278 int dev_change_flags(struct net_device *, unsigned int); 3279 void __dev_notify_flags(struct net_device *, unsigned int old_flags, 3280 unsigned int gchanges); 3281 int dev_change_name(struct net_device *, const char *); 3282 int dev_set_alias(struct net_device *, const char *, size_t); 3283 int dev_get_alias(const struct net_device *, char *, size_t); 3284 int dev_change_net_namespace(struct net_device *, struct net *, const char *); 3285 int __dev_set_mtu(struct net_device *, int); 3286 int dev_set_mtu(struct net_device *, int); 3287 void dev_set_group(struct net_device *, int); 3288 int dev_set_mac_address(struct net_device *, struct sockaddr *); 3289 int dev_change_carrier(struct net_device *, bool new_carrier); 3290 int dev_get_phys_port_id(struct net_device *dev, 3291 struct netdev_phys_item_id *ppid); 3292 int dev_get_phys_port_name(struct net_device *dev, 3293 char *name, size_t len); 3294 int dev_change_proto_down(struct net_device *dev, bool proto_down); 3295 struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev); 3296 struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev, 3297 struct netdev_queue *txq, int *ret); 3298 3299 typedef int (*xdp_op_t)(struct net_device *dev, struct netdev_xdp *xdp); 3300 int dev_change_xdp_fd(struct net_device *dev, struct netlink_ext_ack *extack, 3301 int fd, u32 flags); 3302 u8 __dev_xdp_attached(struct net_device *dev, xdp_op_t xdp_op, u32 *prog_id); 3303 3304 int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb); 3305 int dev_forward_skb(struct net_device *dev, struct sk_buff *skb); 3306 bool is_skb_forwardable(const struct net_device *dev, 3307 const struct sk_buff *skb); 3308 3309 static __always_inline int ____dev_forward_skb(struct net_device *dev, 3310 struct sk_buff *skb) 3311 { 3312 if (skb_orphan_frags(skb, GFP_ATOMIC) || 3313 unlikely(!is_skb_forwardable(dev, skb))) { 3314 atomic_long_inc(&dev->rx_dropped); 3315 kfree_skb(skb); 3316 return NET_RX_DROP; 3317 } 3318 3319 skb_scrub_packet(skb, true); 3320 skb->priority = 0; 3321 return 0; 3322 } 3323 3324 void dev_queue_xmit_nit(struct sk_buff *skb, struct net_device *dev); 3325 3326 extern int netdev_budget; 3327 extern unsigned int netdev_budget_usecs; 3328 3329 /* Called by rtnetlink.c:rtnl_unlock() */ 3330 void netdev_run_todo(void); 3331 3332 /** 3333 * dev_put - release reference to device 3334 * @dev: network device 3335 * 3336 * Release reference to device to allow it to be freed. 3337 */ 3338 static inline void dev_put(struct net_device *dev) 3339 { 3340 this_cpu_dec(*dev->pcpu_refcnt); 3341 } 3342 3343 /** 3344 * dev_hold - get reference to device 3345 * @dev: network device 3346 * 3347 * Hold reference to device to keep it from being freed. 3348 */ 3349 static inline void dev_hold(struct net_device *dev) 3350 { 3351 this_cpu_inc(*dev->pcpu_refcnt); 3352 } 3353 3354 /* Carrier loss detection, dial on demand. The functions netif_carrier_on 3355 * and _off may be called from IRQ context, but it is caller 3356 * who is responsible for serialization of these calls. 3357 * 3358 * The name carrier is inappropriate, these functions should really be 3359 * called netif_lowerlayer_*() because they represent the state of any 3360 * kind of lower layer not just hardware media. 3361 */ 3362 3363 void linkwatch_init_dev(struct net_device *dev); 3364 void linkwatch_fire_event(struct net_device *dev); 3365 void linkwatch_forget_dev(struct net_device *dev); 3366 3367 /** 3368 * netif_carrier_ok - test if carrier present 3369 * @dev: network device 3370 * 3371 * Check if carrier is present on device 3372 */ 3373 static inline bool netif_carrier_ok(const struct net_device *dev) 3374 { 3375 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state); 3376 } 3377 3378 unsigned long dev_trans_start(struct net_device *dev); 3379 3380 void __netdev_watchdog_up(struct net_device *dev); 3381 3382 void netif_carrier_on(struct net_device *dev); 3383 3384 void netif_carrier_off(struct net_device *dev); 3385 3386 /** 3387 * netif_dormant_on - mark device as dormant. 3388 * @dev: network device 3389 * 3390 * Mark device as dormant (as per RFC2863). 3391 * 3392 * The dormant state indicates that the relevant interface is not 3393 * actually in a condition to pass packets (i.e., it is not 'up') but is 3394 * in a "pending" state, waiting for some external event. For "on- 3395 * demand" interfaces, this new state identifies the situation where the 3396 * interface is waiting for events to place it in the up state. 3397 */ 3398 static inline void netif_dormant_on(struct net_device *dev) 3399 { 3400 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state)) 3401 linkwatch_fire_event(dev); 3402 } 3403 3404 /** 3405 * netif_dormant_off - set device as not dormant. 3406 * @dev: network device 3407 * 3408 * Device is not in dormant state. 3409 */ 3410 static inline void netif_dormant_off(struct net_device *dev) 3411 { 3412 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state)) 3413 linkwatch_fire_event(dev); 3414 } 3415 3416 /** 3417 * netif_dormant - test if device is dormant 3418 * @dev: network device 3419 * 3420 * Check if device is dormant. 3421 */ 3422 static inline bool netif_dormant(const struct net_device *dev) 3423 { 3424 return test_bit(__LINK_STATE_DORMANT, &dev->state); 3425 } 3426 3427 3428 /** 3429 * netif_oper_up - test if device is operational 3430 * @dev: network device 3431 * 3432 * Check if carrier is operational 3433 */ 3434 static inline bool netif_oper_up(const struct net_device *dev) 3435 { 3436 return (dev->operstate == IF_OPER_UP || 3437 dev->operstate == IF_OPER_UNKNOWN /* backward compat */); 3438 } 3439 3440 /** 3441 * netif_device_present - is device available or removed 3442 * @dev: network device 3443 * 3444 * Check if device has not been removed from system. 3445 */ 3446 static inline bool netif_device_present(struct net_device *dev) 3447 { 3448 return test_bit(__LINK_STATE_PRESENT, &dev->state); 3449 } 3450 3451 void netif_device_detach(struct net_device *dev); 3452 3453 void netif_device_attach(struct net_device *dev); 3454 3455 /* 3456 * Network interface message level settings 3457 */ 3458 3459 enum { 3460 NETIF_MSG_DRV = 0x0001, 3461 NETIF_MSG_PROBE = 0x0002, 3462 NETIF_MSG_LINK = 0x0004, 3463 NETIF_MSG_TIMER = 0x0008, 3464 NETIF_MSG_IFDOWN = 0x0010, 3465 NETIF_MSG_IFUP = 0x0020, 3466 NETIF_MSG_RX_ERR = 0x0040, 3467 NETIF_MSG_TX_ERR = 0x0080, 3468 NETIF_MSG_TX_QUEUED = 0x0100, 3469 NETIF_MSG_INTR = 0x0200, 3470 NETIF_MSG_TX_DONE = 0x0400, 3471 NETIF_MSG_RX_STATUS = 0x0800, 3472 NETIF_MSG_PKTDATA = 0x1000, 3473 NETIF_MSG_HW = 0x2000, 3474 NETIF_MSG_WOL = 0x4000, 3475 }; 3476 3477 #define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV) 3478 #define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE) 3479 #define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK) 3480 #define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER) 3481 #define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN) 3482 #define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP) 3483 #define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR) 3484 #define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR) 3485 #define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED) 3486 #define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR) 3487 #define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE) 3488 #define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS) 3489 #define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA) 3490 #define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW) 3491 #define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL) 3492 3493 static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits) 3494 { 3495 /* use default */ 3496 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8)) 3497 return default_msg_enable_bits; 3498 if (debug_value == 0) /* no output */ 3499 return 0; 3500 /* set low N bits */ 3501 return (1 << debug_value) - 1; 3502 } 3503 3504 static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu) 3505 { 3506 spin_lock(&txq->_xmit_lock); 3507 txq->xmit_lock_owner = cpu; 3508 } 3509 3510 static inline bool __netif_tx_acquire(struct netdev_queue *txq) 3511 { 3512 __acquire(&txq->_xmit_lock); 3513 return true; 3514 } 3515 3516 static inline void __netif_tx_release(struct netdev_queue *txq) 3517 { 3518 __release(&txq->_xmit_lock); 3519 } 3520 3521 static inline void __netif_tx_lock_bh(struct netdev_queue *txq) 3522 { 3523 spin_lock_bh(&txq->_xmit_lock); 3524 txq->xmit_lock_owner = smp_processor_id(); 3525 } 3526 3527 static inline bool __netif_tx_trylock(struct netdev_queue *txq) 3528 { 3529 bool ok = spin_trylock(&txq->_xmit_lock); 3530 if (likely(ok)) 3531 txq->xmit_lock_owner = smp_processor_id(); 3532 return ok; 3533 } 3534 3535 static inline void __netif_tx_unlock(struct netdev_queue *txq) 3536 { 3537 txq->xmit_lock_owner = -1; 3538 spin_unlock(&txq->_xmit_lock); 3539 } 3540 3541 static inline void __netif_tx_unlock_bh(struct netdev_queue *txq) 3542 { 3543 txq->xmit_lock_owner = -1; 3544 spin_unlock_bh(&txq->_xmit_lock); 3545 } 3546 3547 static inline void txq_trans_update(struct netdev_queue *txq) 3548 { 3549 if (txq->xmit_lock_owner != -1) 3550 txq->trans_start = jiffies; 3551 } 3552 3553 /* legacy drivers only, netdev_start_xmit() sets txq->trans_start */ 3554 static inline void netif_trans_update(struct net_device *dev) 3555 { 3556 struct netdev_queue *txq = netdev_get_tx_queue(dev, 0); 3557 3558 if (txq->trans_start != jiffies) 3559 txq->trans_start = jiffies; 3560 } 3561 3562 /** 3563 * netif_tx_lock - grab network device transmit lock 3564 * @dev: network device 3565 * 3566 * Get network device transmit lock 3567 */ 3568 static inline void netif_tx_lock(struct net_device *dev) 3569 { 3570 unsigned int i; 3571 int cpu; 3572 3573 spin_lock(&dev->tx_global_lock); 3574 cpu = smp_processor_id(); 3575 for (i = 0; i < dev->num_tx_queues; i++) { 3576 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 3577 3578 /* We are the only thread of execution doing a 3579 * freeze, but we have to grab the _xmit_lock in 3580 * order to synchronize with threads which are in 3581 * the ->hard_start_xmit() handler and already 3582 * checked the frozen bit. 3583 */ 3584 __netif_tx_lock(txq, cpu); 3585 set_bit(__QUEUE_STATE_FROZEN, &txq->state); 3586 __netif_tx_unlock(txq); 3587 } 3588 } 3589 3590 static inline void netif_tx_lock_bh(struct net_device *dev) 3591 { 3592 local_bh_disable(); 3593 netif_tx_lock(dev); 3594 } 3595 3596 static inline void netif_tx_unlock(struct net_device *dev) 3597 { 3598 unsigned int i; 3599 3600 for (i = 0; i < dev->num_tx_queues; i++) { 3601 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 3602 3603 /* No need to grab the _xmit_lock here. If the 3604 * queue is not stopped for another reason, we 3605 * force a schedule. 3606 */ 3607 clear_bit(__QUEUE_STATE_FROZEN, &txq->state); 3608 netif_schedule_queue(txq); 3609 } 3610 spin_unlock(&dev->tx_global_lock); 3611 } 3612 3613 static inline void netif_tx_unlock_bh(struct net_device *dev) 3614 { 3615 netif_tx_unlock(dev); 3616 local_bh_enable(); 3617 } 3618 3619 #define HARD_TX_LOCK(dev, txq, cpu) { \ 3620 if ((dev->features & NETIF_F_LLTX) == 0) { \ 3621 __netif_tx_lock(txq, cpu); \ 3622 } else { \ 3623 __netif_tx_acquire(txq); \ 3624 } \ 3625 } 3626 3627 #define HARD_TX_TRYLOCK(dev, txq) \ 3628 (((dev->features & NETIF_F_LLTX) == 0) ? \ 3629 __netif_tx_trylock(txq) : \ 3630 __netif_tx_acquire(txq)) 3631 3632 #define HARD_TX_UNLOCK(dev, txq) { \ 3633 if ((dev->features & NETIF_F_LLTX) == 0) { \ 3634 __netif_tx_unlock(txq); \ 3635 } else { \ 3636 __netif_tx_release(txq); \ 3637 } \ 3638 } 3639 3640 static inline void netif_tx_disable(struct net_device *dev) 3641 { 3642 unsigned int i; 3643 int cpu; 3644 3645 local_bh_disable(); 3646 cpu = smp_processor_id(); 3647 for (i = 0; i < dev->num_tx_queues; i++) { 3648 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 3649 3650 __netif_tx_lock(txq, cpu); 3651 netif_tx_stop_queue(txq); 3652 __netif_tx_unlock(txq); 3653 } 3654 local_bh_enable(); 3655 } 3656 3657 static inline void netif_addr_lock(struct net_device *dev) 3658 { 3659 spin_lock(&dev->addr_list_lock); 3660 } 3661 3662 static inline void netif_addr_lock_nested(struct net_device *dev) 3663 { 3664 int subclass = SINGLE_DEPTH_NESTING; 3665 3666 if (dev->netdev_ops->ndo_get_lock_subclass) 3667 subclass = dev->netdev_ops->ndo_get_lock_subclass(dev); 3668 3669 spin_lock_nested(&dev->addr_list_lock, subclass); 3670 } 3671 3672 static inline void netif_addr_lock_bh(struct net_device *dev) 3673 { 3674 spin_lock_bh(&dev->addr_list_lock); 3675 } 3676 3677 static inline void netif_addr_unlock(struct net_device *dev) 3678 { 3679 spin_unlock(&dev->addr_list_lock); 3680 } 3681 3682 static inline void netif_addr_unlock_bh(struct net_device *dev) 3683 { 3684 spin_unlock_bh(&dev->addr_list_lock); 3685 } 3686 3687 /* 3688 * dev_addrs walker. Should be used only for read access. Call with 3689 * rcu_read_lock held. 3690 */ 3691 #define for_each_dev_addr(dev, ha) \ 3692 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list) 3693 3694 /* These functions live elsewhere (drivers/net/net_init.c, but related) */ 3695 3696 void ether_setup(struct net_device *dev); 3697 3698 /* Support for loadable net-drivers */ 3699 struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name, 3700 unsigned char name_assign_type, 3701 void (*setup)(struct net_device *), 3702 unsigned int txqs, unsigned int rxqs); 3703 #define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \ 3704 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1) 3705 3706 #define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \ 3707 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \ 3708 count) 3709 3710 int register_netdev(struct net_device *dev); 3711 void unregister_netdev(struct net_device *dev); 3712 3713 /* General hardware address lists handling functions */ 3714 int __hw_addr_sync(struct netdev_hw_addr_list *to_list, 3715 struct netdev_hw_addr_list *from_list, int addr_len); 3716 void __hw_addr_unsync(struct netdev_hw_addr_list *to_list, 3717 struct netdev_hw_addr_list *from_list, int addr_len); 3718 int __hw_addr_sync_dev(struct netdev_hw_addr_list *list, 3719 struct net_device *dev, 3720 int (*sync)(struct net_device *, const unsigned char *), 3721 int (*unsync)(struct net_device *, 3722 const unsigned char *)); 3723 void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list, 3724 struct net_device *dev, 3725 int (*unsync)(struct net_device *, 3726 const unsigned char *)); 3727 void __hw_addr_init(struct netdev_hw_addr_list *list); 3728 3729 /* Functions used for device addresses handling */ 3730 int dev_addr_add(struct net_device *dev, const unsigned char *addr, 3731 unsigned char addr_type); 3732 int dev_addr_del(struct net_device *dev, const unsigned char *addr, 3733 unsigned char addr_type); 3734 void dev_addr_flush(struct net_device *dev); 3735 int dev_addr_init(struct net_device *dev); 3736 3737 /* Functions used for unicast addresses handling */ 3738 int dev_uc_add(struct net_device *dev, const unsigned char *addr); 3739 int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr); 3740 int dev_uc_del(struct net_device *dev, const unsigned char *addr); 3741 int dev_uc_sync(struct net_device *to, struct net_device *from); 3742 int dev_uc_sync_multiple(struct net_device *to, struct net_device *from); 3743 void dev_uc_unsync(struct net_device *to, struct net_device *from); 3744 void dev_uc_flush(struct net_device *dev); 3745 void dev_uc_init(struct net_device *dev); 3746 3747 /** 3748 * __dev_uc_sync - Synchonize device's unicast list 3749 * @dev: device to sync 3750 * @sync: function to call if address should be added 3751 * @unsync: function to call if address should be removed 3752 * 3753 * Add newly added addresses to the interface, and release 3754 * addresses that have been deleted. 3755 */ 3756 static inline int __dev_uc_sync(struct net_device *dev, 3757 int (*sync)(struct net_device *, 3758 const unsigned char *), 3759 int (*unsync)(struct net_device *, 3760 const unsigned char *)) 3761 { 3762 return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync); 3763 } 3764 3765 /** 3766 * __dev_uc_unsync - Remove synchronized addresses from device 3767 * @dev: device to sync 3768 * @unsync: function to call if address should be removed 3769 * 3770 * Remove all addresses that were added to the device by dev_uc_sync(). 3771 */ 3772 static inline void __dev_uc_unsync(struct net_device *dev, 3773 int (*unsync)(struct net_device *, 3774 const unsigned char *)) 3775 { 3776 __hw_addr_unsync_dev(&dev->uc, dev, unsync); 3777 } 3778 3779 /* Functions used for multicast addresses handling */ 3780 int dev_mc_add(struct net_device *dev, const unsigned char *addr); 3781 int dev_mc_add_global(struct net_device *dev, const unsigned char *addr); 3782 int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr); 3783 int dev_mc_del(struct net_device *dev, const unsigned char *addr); 3784 int dev_mc_del_global(struct net_device *dev, const unsigned char *addr); 3785 int dev_mc_sync(struct net_device *to, struct net_device *from); 3786 int dev_mc_sync_multiple(struct net_device *to, struct net_device *from); 3787 void dev_mc_unsync(struct net_device *to, struct net_device *from); 3788 void dev_mc_flush(struct net_device *dev); 3789 void dev_mc_init(struct net_device *dev); 3790 3791 /** 3792 * __dev_mc_sync - Synchonize device's multicast list 3793 * @dev: device to sync 3794 * @sync: function to call if address should be added 3795 * @unsync: function to call if address should be removed 3796 * 3797 * Add newly added addresses to the interface, and release 3798 * addresses that have been deleted. 3799 */ 3800 static inline int __dev_mc_sync(struct net_device *dev, 3801 int (*sync)(struct net_device *, 3802 const unsigned char *), 3803 int (*unsync)(struct net_device *, 3804 const unsigned char *)) 3805 { 3806 return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync); 3807 } 3808 3809 /** 3810 * __dev_mc_unsync - Remove synchronized addresses from device 3811 * @dev: device to sync 3812 * @unsync: function to call if address should be removed 3813 * 3814 * Remove all addresses that were added to the device by dev_mc_sync(). 3815 */ 3816 static inline void __dev_mc_unsync(struct net_device *dev, 3817 int (*unsync)(struct net_device *, 3818 const unsigned char *)) 3819 { 3820 __hw_addr_unsync_dev(&dev->mc, dev, unsync); 3821 } 3822 3823 /* Functions used for secondary unicast and multicast support */ 3824 void dev_set_rx_mode(struct net_device *dev); 3825 void __dev_set_rx_mode(struct net_device *dev); 3826 int dev_set_promiscuity(struct net_device *dev, int inc); 3827 int dev_set_allmulti(struct net_device *dev, int inc); 3828 void netdev_state_change(struct net_device *dev); 3829 void netdev_notify_peers(struct net_device *dev); 3830 void netdev_features_change(struct net_device *dev); 3831 /* Load a device via the kmod */ 3832 void dev_load(struct net *net, const char *name); 3833 struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev, 3834 struct rtnl_link_stats64 *storage); 3835 void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64, 3836 const struct net_device_stats *netdev_stats); 3837 3838 extern int netdev_max_backlog; 3839 extern int netdev_tstamp_prequeue; 3840 extern int weight_p; 3841 extern int dev_weight_rx_bias; 3842 extern int dev_weight_tx_bias; 3843 extern int dev_rx_weight; 3844 extern int dev_tx_weight; 3845 3846 bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev); 3847 struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev, 3848 struct list_head **iter); 3849 struct net_device *netdev_all_upper_get_next_dev_rcu(struct net_device *dev, 3850 struct list_head **iter); 3851 3852 /* iterate through upper list, must be called under RCU read lock */ 3853 #define netdev_for_each_upper_dev_rcu(dev, updev, iter) \ 3854 for (iter = &(dev)->adj_list.upper, \ 3855 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \ 3856 updev; \ 3857 updev = netdev_upper_get_next_dev_rcu(dev, &(iter))) 3858 3859 int netdev_walk_all_upper_dev_rcu(struct net_device *dev, 3860 int (*fn)(struct net_device *upper_dev, 3861 void *data), 3862 void *data); 3863 3864 bool netdev_has_upper_dev_all_rcu(struct net_device *dev, 3865 struct net_device *upper_dev); 3866 3867 bool netdev_has_any_upper_dev(struct net_device *dev); 3868 3869 void *netdev_lower_get_next_private(struct net_device *dev, 3870 struct list_head **iter); 3871 void *netdev_lower_get_next_private_rcu(struct net_device *dev, 3872 struct list_head **iter); 3873 3874 #define netdev_for_each_lower_private(dev, priv, iter) \ 3875 for (iter = (dev)->adj_list.lower.next, \ 3876 priv = netdev_lower_get_next_private(dev, &(iter)); \ 3877 priv; \ 3878 priv = netdev_lower_get_next_private(dev, &(iter))) 3879 3880 #define netdev_for_each_lower_private_rcu(dev, priv, iter) \ 3881 for (iter = &(dev)->adj_list.lower, \ 3882 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \ 3883 priv; \ 3884 priv = netdev_lower_get_next_private_rcu(dev, &(iter))) 3885 3886 void *netdev_lower_get_next(struct net_device *dev, 3887 struct list_head **iter); 3888 3889 #define netdev_for_each_lower_dev(dev, ldev, iter) \ 3890 for (iter = (dev)->adj_list.lower.next, \ 3891 ldev = netdev_lower_get_next(dev, &(iter)); \ 3892 ldev; \ 3893 ldev = netdev_lower_get_next(dev, &(iter))) 3894 3895 struct net_device *netdev_all_lower_get_next(struct net_device *dev, 3896 struct list_head **iter); 3897 struct net_device *netdev_all_lower_get_next_rcu(struct net_device *dev, 3898 struct list_head **iter); 3899 3900 int netdev_walk_all_lower_dev(struct net_device *dev, 3901 int (*fn)(struct net_device *lower_dev, 3902 void *data), 3903 void *data); 3904 int netdev_walk_all_lower_dev_rcu(struct net_device *dev, 3905 int (*fn)(struct net_device *lower_dev, 3906 void *data), 3907 void *data); 3908 3909 void *netdev_adjacent_get_private(struct list_head *adj_list); 3910 void *netdev_lower_get_first_private_rcu(struct net_device *dev); 3911 struct net_device *netdev_master_upper_dev_get(struct net_device *dev); 3912 struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev); 3913 int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev); 3914 int netdev_master_upper_dev_link(struct net_device *dev, 3915 struct net_device *upper_dev, 3916 void *upper_priv, void *upper_info); 3917 void netdev_upper_dev_unlink(struct net_device *dev, 3918 struct net_device *upper_dev); 3919 void netdev_adjacent_rename_links(struct net_device *dev, char *oldname); 3920 void *netdev_lower_dev_get_private(struct net_device *dev, 3921 struct net_device *lower_dev); 3922 void netdev_lower_state_changed(struct net_device *lower_dev, 3923 void *lower_state_info); 3924 3925 /* RSS keys are 40 or 52 bytes long */ 3926 #define NETDEV_RSS_KEY_LEN 52 3927 extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN] __read_mostly; 3928 void netdev_rss_key_fill(void *buffer, size_t len); 3929 3930 int dev_get_nest_level(struct net_device *dev); 3931 int skb_checksum_help(struct sk_buff *skb); 3932 int skb_crc32c_csum_help(struct sk_buff *skb); 3933 int skb_csum_hwoffload_help(struct sk_buff *skb, 3934 const netdev_features_t features); 3935 3936 struct sk_buff *__skb_gso_segment(struct sk_buff *skb, 3937 netdev_features_t features, bool tx_path); 3938 struct sk_buff *skb_mac_gso_segment(struct sk_buff *skb, 3939 netdev_features_t features); 3940 3941 struct netdev_bonding_info { 3942 ifslave slave; 3943 ifbond master; 3944 }; 3945 3946 struct netdev_notifier_bonding_info { 3947 struct netdev_notifier_info info; /* must be first */ 3948 struct netdev_bonding_info bonding_info; 3949 }; 3950 3951 void netdev_bonding_info_change(struct net_device *dev, 3952 struct netdev_bonding_info *bonding_info); 3953 3954 static inline 3955 struct sk_buff *skb_gso_segment(struct sk_buff *skb, netdev_features_t features) 3956 { 3957 return __skb_gso_segment(skb, features, true); 3958 } 3959 __be16 skb_network_protocol(struct sk_buff *skb, int *depth); 3960 3961 static inline bool can_checksum_protocol(netdev_features_t features, 3962 __be16 protocol) 3963 { 3964 if (protocol == htons(ETH_P_FCOE)) 3965 return !!(features & NETIF_F_FCOE_CRC); 3966 3967 /* Assume this is an IP checksum (not SCTP CRC) */ 3968 3969 if (features & NETIF_F_HW_CSUM) { 3970 /* Can checksum everything */ 3971 return true; 3972 } 3973 3974 switch (protocol) { 3975 case htons(ETH_P_IP): 3976 return !!(features & NETIF_F_IP_CSUM); 3977 case htons(ETH_P_IPV6): 3978 return !!(features & NETIF_F_IPV6_CSUM); 3979 default: 3980 return false; 3981 } 3982 } 3983 3984 #ifdef CONFIG_BUG 3985 void netdev_rx_csum_fault(struct net_device *dev); 3986 #else 3987 static inline void netdev_rx_csum_fault(struct net_device *dev) 3988 { 3989 } 3990 #endif 3991 /* rx skb timestamps */ 3992 void net_enable_timestamp(void); 3993 void net_disable_timestamp(void); 3994 3995 #ifdef CONFIG_PROC_FS 3996 int __init dev_proc_init(void); 3997 #else 3998 #define dev_proc_init() 0 3999 #endif 4000 4001 static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops, 4002 struct sk_buff *skb, struct net_device *dev, 4003 bool more) 4004 { 4005 skb->xmit_more = more ? 1 : 0; 4006 return ops->ndo_start_xmit(skb, dev); 4007 } 4008 4009 static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev, 4010 struct netdev_queue *txq, bool more) 4011 { 4012 const struct net_device_ops *ops = dev->netdev_ops; 4013 int rc; 4014 4015 rc = __netdev_start_xmit(ops, skb, dev, more); 4016 if (rc == NETDEV_TX_OK) 4017 txq_trans_update(txq); 4018 4019 return rc; 4020 } 4021 4022 int netdev_class_create_file_ns(const struct class_attribute *class_attr, 4023 const void *ns); 4024 void netdev_class_remove_file_ns(const struct class_attribute *class_attr, 4025 const void *ns); 4026 4027 static inline int netdev_class_create_file(const struct class_attribute *class_attr) 4028 { 4029 return netdev_class_create_file_ns(class_attr, NULL); 4030 } 4031 4032 static inline void netdev_class_remove_file(const struct class_attribute *class_attr) 4033 { 4034 netdev_class_remove_file_ns(class_attr, NULL); 4035 } 4036 4037 extern const struct kobj_ns_type_operations net_ns_type_operations; 4038 4039 const char *netdev_drivername(const struct net_device *dev); 4040 4041 void linkwatch_run_queue(void); 4042 4043 static inline netdev_features_t netdev_intersect_features(netdev_features_t f1, 4044 netdev_features_t f2) 4045 { 4046 if ((f1 ^ f2) & NETIF_F_HW_CSUM) { 4047 if (f1 & NETIF_F_HW_CSUM) 4048 f1 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM); 4049 else 4050 f2 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM); 4051 } 4052 4053 return f1 & f2; 4054 } 4055 4056 static inline netdev_features_t netdev_get_wanted_features( 4057 struct net_device *dev) 4058 { 4059 return (dev->features & ~dev->hw_features) | dev->wanted_features; 4060 } 4061 netdev_features_t netdev_increment_features(netdev_features_t all, 4062 netdev_features_t one, netdev_features_t mask); 4063 4064 /* Allow TSO being used on stacked device : 4065 * Performing the GSO segmentation before last device 4066 * is a performance improvement. 4067 */ 4068 static inline netdev_features_t netdev_add_tso_features(netdev_features_t features, 4069 netdev_features_t mask) 4070 { 4071 return netdev_increment_features(features, NETIF_F_ALL_TSO, mask); 4072 } 4073 4074 int __netdev_update_features(struct net_device *dev); 4075 void netdev_update_features(struct net_device *dev); 4076 void netdev_change_features(struct net_device *dev); 4077 4078 void netif_stacked_transfer_operstate(const struct net_device *rootdev, 4079 struct net_device *dev); 4080 4081 netdev_features_t passthru_features_check(struct sk_buff *skb, 4082 struct net_device *dev, 4083 netdev_features_t features); 4084 netdev_features_t netif_skb_features(struct sk_buff *skb); 4085 4086 static inline bool net_gso_ok(netdev_features_t features, int gso_type) 4087 { 4088 netdev_features_t feature = (netdev_features_t)gso_type << NETIF_F_GSO_SHIFT; 4089 4090 /* check flags correspondence */ 4091 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT)); 4092 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT)); 4093 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT)); 4094 BUILD_BUG_ON(SKB_GSO_TCP_FIXEDID != (NETIF_F_TSO_MANGLEID >> NETIF_F_GSO_SHIFT)); 4095 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT)); 4096 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT)); 4097 BUILD_BUG_ON(SKB_GSO_GRE != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT)); 4098 BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT)); 4099 BUILD_BUG_ON(SKB_GSO_IPXIP4 != (NETIF_F_GSO_IPXIP4 >> NETIF_F_GSO_SHIFT)); 4100 BUILD_BUG_ON(SKB_GSO_IPXIP6 != (NETIF_F_GSO_IPXIP6 >> NETIF_F_GSO_SHIFT)); 4101 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT)); 4102 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT)); 4103 BUILD_BUG_ON(SKB_GSO_PARTIAL != (NETIF_F_GSO_PARTIAL >> NETIF_F_GSO_SHIFT)); 4104 BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT)); 4105 BUILD_BUG_ON(SKB_GSO_SCTP != (NETIF_F_GSO_SCTP >> NETIF_F_GSO_SHIFT)); 4106 BUILD_BUG_ON(SKB_GSO_ESP != (NETIF_F_GSO_ESP >> NETIF_F_GSO_SHIFT)); 4107 4108 return (features & feature) == feature; 4109 } 4110 4111 static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features) 4112 { 4113 return net_gso_ok(features, skb_shinfo(skb)->gso_type) && 4114 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST)); 4115 } 4116 4117 static inline bool netif_needs_gso(struct sk_buff *skb, 4118 netdev_features_t features) 4119 { 4120 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) || 4121 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) && 4122 (skb->ip_summed != CHECKSUM_UNNECESSARY))); 4123 } 4124 4125 static inline void netif_set_gso_max_size(struct net_device *dev, 4126 unsigned int size) 4127 { 4128 dev->gso_max_size = size; 4129 } 4130 4131 static inline void skb_gso_error_unwind(struct sk_buff *skb, __be16 protocol, 4132 int pulled_hlen, u16 mac_offset, 4133 int mac_len) 4134 { 4135 skb->protocol = protocol; 4136 skb->encapsulation = 1; 4137 skb_push(skb, pulled_hlen); 4138 skb_reset_transport_header(skb); 4139 skb->mac_header = mac_offset; 4140 skb->network_header = skb->mac_header + mac_len; 4141 skb->mac_len = mac_len; 4142 } 4143 4144 static inline bool netif_is_macsec(const struct net_device *dev) 4145 { 4146 return dev->priv_flags & IFF_MACSEC; 4147 } 4148 4149 static inline bool netif_is_macvlan(const struct net_device *dev) 4150 { 4151 return dev->priv_flags & IFF_MACVLAN; 4152 } 4153 4154 static inline bool netif_is_macvlan_port(const struct net_device *dev) 4155 { 4156 return dev->priv_flags & IFF_MACVLAN_PORT; 4157 } 4158 4159 static inline bool netif_is_ipvlan(const struct net_device *dev) 4160 { 4161 return dev->priv_flags & IFF_IPVLAN_SLAVE; 4162 } 4163 4164 static inline bool netif_is_ipvlan_port(const struct net_device *dev) 4165 { 4166 return dev->priv_flags & IFF_IPVLAN_MASTER; 4167 } 4168 4169 static inline bool netif_is_bond_master(const struct net_device *dev) 4170 { 4171 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING; 4172 } 4173 4174 static inline bool netif_is_bond_slave(const struct net_device *dev) 4175 { 4176 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING; 4177 } 4178 4179 static inline bool netif_supports_nofcs(struct net_device *dev) 4180 { 4181 return dev->priv_flags & IFF_SUPP_NOFCS; 4182 } 4183 4184 static inline bool netif_is_l3_master(const struct net_device *dev) 4185 { 4186 return dev->priv_flags & IFF_L3MDEV_MASTER; 4187 } 4188 4189 static inline bool netif_is_l3_slave(const struct net_device *dev) 4190 { 4191 return dev->priv_flags & IFF_L3MDEV_SLAVE; 4192 } 4193 4194 static inline bool netif_is_bridge_master(const struct net_device *dev) 4195 { 4196 return dev->priv_flags & IFF_EBRIDGE; 4197 } 4198 4199 static inline bool netif_is_bridge_port(const struct net_device *dev) 4200 { 4201 return dev->priv_flags & IFF_BRIDGE_PORT; 4202 } 4203 4204 static inline bool netif_is_ovs_master(const struct net_device *dev) 4205 { 4206 return dev->priv_flags & IFF_OPENVSWITCH; 4207 } 4208 4209 static inline bool netif_is_ovs_port(const struct net_device *dev) 4210 { 4211 return dev->priv_flags & IFF_OVS_DATAPATH; 4212 } 4213 4214 static inline bool netif_is_team_master(const struct net_device *dev) 4215 { 4216 return dev->priv_flags & IFF_TEAM; 4217 } 4218 4219 static inline bool netif_is_team_port(const struct net_device *dev) 4220 { 4221 return dev->priv_flags & IFF_TEAM_PORT; 4222 } 4223 4224 static inline bool netif_is_lag_master(const struct net_device *dev) 4225 { 4226 return netif_is_bond_master(dev) || netif_is_team_master(dev); 4227 } 4228 4229 static inline bool netif_is_lag_port(const struct net_device *dev) 4230 { 4231 return netif_is_bond_slave(dev) || netif_is_team_port(dev); 4232 } 4233 4234 static inline bool netif_is_rxfh_configured(const struct net_device *dev) 4235 { 4236 return dev->priv_flags & IFF_RXFH_CONFIGURED; 4237 } 4238 4239 /* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */ 4240 static inline void netif_keep_dst(struct net_device *dev) 4241 { 4242 dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM); 4243 } 4244 4245 /* return true if dev can't cope with mtu frames that need vlan tag insertion */ 4246 static inline bool netif_reduces_vlan_mtu(struct net_device *dev) 4247 { 4248 /* TODO: reserve and use an additional IFF bit, if we get more users */ 4249 return dev->priv_flags & IFF_MACSEC; 4250 } 4251 4252 extern struct pernet_operations __net_initdata loopback_net_ops; 4253 4254 /* Logging, debugging and troubleshooting/diagnostic helpers. */ 4255 4256 /* netdev_printk helpers, similar to dev_printk */ 4257 4258 static inline const char *netdev_name(const struct net_device *dev) 4259 { 4260 if (!dev->name[0] || strchr(dev->name, '%')) 4261 return "(unnamed net_device)"; 4262 return dev->name; 4263 } 4264 4265 static inline bool netdev_unregistering(const struct net_device *dev) 4266 { 4267 return dev->reg_state == NETREG_UNREGISTERING; 4268 } 4269 4270 static inline const char *netdev_reg_state(const struct net_device *dev) 4271 { 4272 switch (dev->reg_state) { 4273 case NETREG_UNINITIALIZED: return " (uninitialized)"; 4274 case NETREG_REGISTERED: return ""; 4275 case NETREG_UNREGISTERING: return " (unregistering)"; 4276 case NETREG_UNREGISTERED: return " (unregistered)"; 4277 case NETREG_RELEASED: return " (released)"; 4278 case NETREG_DUMMY: return " (dummy)"; 4279 } 4280 4281 WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, dev->reg_state); 4282 return " (unknown)"; 4283 } 4284 4285 __printf(3, 4) 4286 void netdev_printk(const char *level, const struct net_device *dev, 4287 const char *format, ...); 4288 __printf(2, 3) 4289 void netdev_emerg(const struct net_device *dev, const char *format, ...); 4290 __printf(2, 3) 4291 void netdev_alert(const struct net_device *dev, const char *format, ...); 4292 __printf(2, 3) 4293 void netdev_crit(const struct net_device *dev, const char *format, ...); 4294 __printf(2, 3) 4295 void netdev_err(const struct net_device *dev, const char *format, ...); 4296 __printf(2, 3) 4297 void netdev_warn(const struct net_device *dev, const char *format, ...); 4298 __printf(2, 3) 4299 void netdev_notice(const struct net_device *dev, const char *format, ...); 4300 __printf(2, 3) 4301 void netdev_info(const struct net_device *dev, const char *format, ...); 4302 4303 #define MODULE_ALIAS_NETDEV(device) \ 4304 MODULE_ALIAS("netdev-" device) 4305 4306 #if defined(CONFIG_DYNAMIC_DEBUG) 4307 #define netdev_dbg(__dev, format, args...) \ 4308 do { \ 4309 dynamic_netdev_dbg(__dev, format, ##args); \ 4310 } while (0) 4311 #elif defined(DEBUG) 4312 #define netdev_dbg(__dev, format, args...) \ 4313 netdev_printk(KERN_DEBUG, __dev, format, ##args) 4314 #else 4315 #define netdev_dbg(__dev, format, args...) \ 4316 ({ \ 4317 if (0) \ 4318 netdev_printk(KERN_DEBUG, __dev, format, ##args); \ 4319 }) 4320 #endif 4321 4322 #if defined(VERBOSE_DEBUG) 4323 #define netdev_vdbg netdev_dbg 4324 #else 4325 4326 #define netdev_vdbg(dev, format, args...) \ 4327 ({ \ 4328 if (0) \ 4329 netdev_printk(KERN_DEBUG, dev, format, ##args); \ 4330 0; \ 4331 }) 4332 #endif 4333 4334 /* 4335 * netdev_WARN() acts like dev_printk(), but with the key difference 4336 * of using a WARN/WARN_ON to get the message out, including the 4337 * file/line information and a backtrace. 4338 */ 4339 #define netdev_WARN(dev, format, args...) \ 4340 WARN(1, "netdevice: %s%s\n" format, netdev_name(dev), \ 4341 netdev_reg_state(dev), ##args) 4342 4343 /* netif printk helpers, similar to netdev_printk */ 4344 4345 #define netif_printk(priv, type, level, dev, fmt, args...) \ 4346 do { \ 4347 if (netif_msg_##type(priv)) \ 4348 netdev_printk(level, (dev), fmt, ##args); \ 4349 } while (0) 4350 4351 #define netif_level(level, priv, type, dev, fmt, args...) \ 4352 do { \ 4353 if (netif_msg_##type(priv)) \ 4354 netdev_##level(dev, fmt, ##args); \ 4355 } while (0) 4356 4357 #define netif_emerg(priv, type, dev, fmt, args...) \ 4358 netif_level(emerg, priv, type, dev, fmt, ##args) 4359 #define netif_alert(priv, type, dev, fmt, args...) \ 4360 netif_level(alert, priv, type, dev, fmt, ##args) 4361 #define netif_crit(priv, type, dev, fmt, args...) \ 4362 netif_level(crit, priv, type, dev, fmt, ##args) 4363 #define netif_err(priv, type, dev, fmt, args...) \ 4364 netif_level(err, priv, type, dev, fmt, ##args) 4365 #define netif_warn(priv, type, dev, fmt, args...) \ 4366 netif_level(warn, priv, type, dev, fmt, ##args) 4367 #define netif_notice(priv, type, dev, fmt, args...) \ 4368 netif_level(notice, priv, type, dev, fmt, ##args) 4369 #define netif_info(priv, type, dev, fmt, args...) \ 4370 netif_level(info, priv, type, dev, fmt, ##args) 4371 4372 #if defined(CONFIG_DYNAMIC_DEBUG) 4373 #define netif_dbg(priv, type, netdev, format, args...) \ 4374 do { \ 4375 if (netif_msg_##type(priv)) \ 4376 dynamic_netdev_dbg(netdev, format, ##args); \ 4377 } while (0) 4378 #elif defined(DEBUG) 4379 #define netif_dbg(priv, type, dev, format, args...) \ 4380 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args) 4381 #else 4382 #define netif_dbg(priv, type, dev, format, args...) \ 4383 ({ \ 4384 if (0) \ 4385 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \ 4386 0; \ 4387 }) 4388 #endif 4389 4390 /* if @cond then downgrade to debug, else print at @level */ 4391 #define netif_cond_dbg(priv, type, netdev, cond, level, fmt, args...) \ 4392 do { \ 4393 if (cond) \ 4394 netif_dbg(priv, type, netdev, fmt, ##args); \ 4395 else \ 4396 netif_ ## level(priv, type, netdev, fmt, ##args); \ 4397 } while (0) 4398 4399 #if defined(VERBOSE_DEBUG) 4400 #define netif_vdbg netif_dbg 4401 #else 4402 #define netif_vdbg(priv, type, dev, format, args...) \ 4403 ({ \ 4404 if (0) \ 4405 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \ 4406 0; \ 4407 }) 4408 #endif 4409 4410 /* 4411 * The list of packet types we will receive (as opposed to discard) 4412 * and the routines to invoke. 4413 * 4414 * Why 16. Because with 16 the only overlap we get on a hash of the 4415 * low nibble of the protocol value is RARP/SNAP/X.25. 4416 * 4417 * NOTE: That is no longer true with the addition of VLAN tags. Not 4418 * sure which should go first, but I bet it won't make much 4419 * difference if we are running VLANs. The good news is that 4420 * this protocol won't be in the list unless compiled in, so 4421 * the average user (w/out VLANs) will not be adversely affected. 4422 * --BLG 4423 * 4424 * 0800 IP 4425 * 8100 802.1Q VLAN 4426 * 0001 802.3 4427 * 0002 AX.25 4428 * 0004 802.2 4429 * 8035 RARP 4430 * 0005 SNAP 4431 * 0805 X.25 4432 * 0806 ARP 4433 * 8137 IPX 4434 * 0009 Localtalk 4435 * 86DD IPv6 4436 */ 4437 #define PTYPE_HASH_SIZE (16) 4438 #define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1) 4439 4440 #endif /* _LINUX_NETDEVICE_H */ 4441