1 /* 2 * INET An implementation of the TCP/IP protocol suite for the LINUX 3 * operating system. INET is implemented using the BSD Socket 4 * interface as the means of communication with the user level. 5 * 6 * Definitions for the Interfaces handler. 7 * 8 * Version: @(#)dev.h 1.0.10 08/12/93 9 * 10 * Authors: Ross Biro 11 * Fred N. van Kempen, <[email protected]> 12 * Corey Minyard <[email protected]> 13 * Donald J. Becker, <[email protected]> 14 * Alan Cox, <[email protected]> 15 * Bjorn Ekwall. <[email protected]> 16 * Pekka Riikonen <[email protected]> 17 * 18 * This program is free software; you can redistribute it and/or 19 * modify it under the terms of the GNU General Public License 20 * as published by the Free Software Foundation; either version 21 * 2 of the License, or (at your option) any later version. 22 * 23 * Moved to /usr/include/linux for NET3 24 */ 25 #ifndef _LINUX_NETDEVICE_H 26 #define _LINUX_NETDEVICE_H 27 28 #include <linux/timer.h> 29 #include <linux/bug.h> 30 #include <linux/delay.h> 31 #include <linux/atomic.h> 32 #include <linux/prefetch.h> 33 #include <asm/cache.h> 34 #include <asm/byteorder.h> 35 36 #include <linux/percpu.h> 37 #include <linux/rculist.h> 38 #include <linux/dmaengine.h> 39 #include <linux/workqueue.h> 40 #include <linux/dynamic_queue_limits.h> 41 42 #include <linux/ethtool.h> 43 #include <net/net_namespace.h> 44 #ifdef CONFIG_DCB 45 #include <net/dcbnl.h> 46 #endif 47 #include <net/netprio_cgroup.h> 48 49 #include <linux/netdev_features.h> 50 #include <linux/neighbour.h> 51 #include <uapi/linux/netdevice.h> 52 #include <uapi/linux/if_bonding.h> 53 #include <uapi/linux/pkt_cls.h> 54 #include <linux/hashtable.h> 55 56 struct netpoll_info; 57 struct device; 58 struct phy_device; 59 struct dsa_switch_tree; 60 61 /* 802.11 specific */ 62 struct wireless_dev; 63 /* 802.15.4 specific */ 64 struct wpan_dev; 65 struct mpls_dev; 66 /* UDP Tunnel offloads */ 67 struct udp_tunnel_info; 68 struct bpf_prog; 69 struct xdp_buff; 70 71 void netdev_set_default_ethtool_ops(struct net_device *dev, 72 const struct ethtool_ops *ops); 73 74 /* Backlog congestion levels */ 75 #define NET_RX_SUCCESS 0 /* keep 'em coming, baby */ 76 #define NET_RX_DROP 1 /* packet dropped */ 77 78 /* 79 * Transmit return codes: transmit return codes originate from three different 80 * namespaces: 81 * 82 * - qdisc return codes 83 * - driver transmit return codes 84 * - errno values 85 * 86 * Drivers are allowed to return any one of those in their hard_start_xmit() 87 * function. Real network devices commonly used with qdiscs should only return 88 * the driver transmit return codes though - when qdiscs are used, the actual 89 * transmission happens asynchronously, so the value is not propagated to 90 * higher layers. Virtual network devices transmit synchronously; in this case 91 * the driver transmit return codes are consumed by dev_queue_xmit(), and all 92 * others are propagated to higher layers. 93 */ 94 95 /* qdisc ->enqueue() return codes. */ 96 #define NET_XMIT_SUCCESS 0x00 97 #define NET_XMIT_DROP 0x01 /* skb dropped */ 98 #define NET_XMIT_CN 0x02 /* congestion notification */ 99 #define NET_XMIT_MASK 0x0f /* qdisc flags in net/sch_generic.h */ 100 101 /* NET_XMIT_CN is special. It does not guarantee that this packet is lost. It 102 * indicates that the device will soon be dropping packets, or already drops 103 * some packets of the same priority; prompting us to send less aggressively. */ 104 #define net_xmit_eval(e) ((e) == NET_XMIT_CN ? 0 : (e)) 105 #define net_xmit_errno(e) ((e) != NET_XMIT_CN ? -ENOBUFS : 0) 106 107 /* Driver transmit return codes */ 108 #define NETDEV_TX_MASK 0xf0 109 110 enum netdev_tx { 111 __NETDEV_TX_MIN = INT_MIN, /* make sure enum is signed */ 112 NETDEV_TX_OK = 0x00, /* driver took care of packet */ 113 NETDEV_TX_BUSY = 0x10, /* driver tx path was busy*/ 114 }; 115 typedef enum netdev_tx netdev_tx_t; 116 117 /* 118 * Current order: NETDEV_TX_MASK > NET_XMIT_MASK >= 0 is significant; 119 * hard_start_xmit() return < NET_XMIT_MASK means skb was consumed. 120 */ 121 static inline bool dev_xmit_complete(int rc) 122 { 123 /* 124 * Positive cases with an skb consumed by a driver: 125 * - successful transmission (rc == NETDEV_TX_OK) 126 * - error while transmitting (rc < 0) 127 * - error while queueing to a different device (rc & NET_XMIT_MASK) 128 */ 129 if (likely(rc < NET_XMIT_MASK)) 130 return true; 131 132 return false; 133 } 134 135 /* 136 * Compute the worst-case header length according to the protocols 137 * used. 138 */ 139 140 #if defined(CONFIG_HYPERV_NET) 141 # define LL_MAX_HEADER 128 142 #elif defined(CONFIG_WLAN) || IS_ENABLED(CONFIG_AX25) 143 # if defined(CONFIG_MAC80211_MESH) 144 # define LL_MAX_HEADER 128 145 # else 146 # define LL_MAX_HEADER 96 147 # endif 148 #else 149 # define LL_MAX_HEADER 32 150 #endif 151 152 #if !IS_ENABLED(CONFIG_NET_IPIP) && !IS_ENABLED(CONFIG_NET_IPGRE) && \ 153 !IS_ENABLED(CONFIG_IPV6_SIT) && !IS_ENABLED(CONFIG_IPV6_TUNNEL) 154 #define MAX_HEADER LL_MAX_HEADER 155 #else 156 #define MAX_HEADER (LL_MAX_HEADER + 48) 157 #endif 158 159 /* 160 * Old network device statistics. Fields are native words 161 * (unsigned long) so they can be read and written atomically. 162 */ 163 164 struct net_device_stats { 165 unsigned long rx_packets; 166 unsigned long tx_packets; 167 unsigned long rx_bytes; 168 unsigned long tx_bytes; 169 unsigned long rx_errors; 170 unsigned long tx_errors; 171 unsigned long rx_dropped; 172 unsigned long tx_dropped; 173 unsigned long multicast; 174 unsigned long collisions; 175 unsigned long rx_length_errors; 176 unsigned long rx_over_errors; 177 unsigned long rx_crc_errors; 178 unsigned long rx_frame_errors; 179 unsigned long rx_fifo_errors; 180 unsigned long rx_missed_errors; 181 unsigned long tx_aborted_errors; 182 unsigned long tx_carrier_errors; 183 unsigned long tx_fifo_errors; 184 unsigned long tx_heartbeat_errors; 185 unsigned long tx_window_errors; 186 unsigned long rx_compressed; 187 unsigned long tx_compressed; 188 }; 189 190 191 #include <linux/cache.h> 192 #include <linux/skbuff.h> 193 194 #ifdef CONFIG_RPS 195 #include <linux/static_key.h> 196 extern struct static_key rps_needed; 197 extern struct static_key rfs_needed; 198 #endif 199 200 struct neighbour; 201 struct neigh_parms; 202 struct sk_buff; 203 204 struct netdev_hw_addr { 205 struct list_head list; 206 unsigned char addr[MAX_ADDR_LEN]; 207 unsigned char type; 208 #define NETDEV_HW_ADDR_T_LAN 1 209 #define NETDEV_HW_ADDR_T_SAN 2 210 #define NETDEV_HW_ADDR_T_SLAVE 3 211 #define NETDEV_HW_ADDR_T_UNICAST 4 212 #define NETDEV_HW_ADDR_T_MULTICAST 5 213 bool global_use; 214 int sync_cnt; 215 int refcount; 216 int synced; 217 struct rcu_head rcu_head; 218 }; 219 220 struct netdev_hw_addr_list { 221 struct list_head list; 222 int count; 223 }; 224 225 #define netdev_hw_addr_list_count(l) ((l)->count) 226 #define netdev_hw_addr_list_empty(l) (netdev_hw_addr_list_count(l) == 0) 227 #define netdev_hw_addr_list_for_each(ha, l) \ 228 list_for_each_entry(ha, &(l)->list, list) 229 230 #define netdev_uc_count(dev) netdev_hw_addr_list_count(&(dev)->uc) 231 #define netdev_uc_empty(dev) netdev_hw_addr_list_empty(&(dev)->uc) 232 #define netdev_for_each_uc_addr(ha, dev) \ 233 netdev_hw_addr_list_for_each(ha, &(dev)->uc) 234 235 #define netdev_mc_count(dev) netdev_hw_addr_list_count(&(dev)->mc) 236 #define netdev_mc_empty(dev) netdev_hw_addr_list_empty(&(dev)->mc) 237 #define netdev_for_each_mc_addr(ha, dev) \ 238 netdev_hw_addr_list_for_each(ha, &(dev)->mc) 239 240 struct hh_cache { 241 unsigned int hh_len; 242 seqlock_t hh_lock; 243 244 /* cached hardware header; allow for machine alignment needs. */ 245 #define HH_DATA_MOD 16 246 #define HH_DATA_OFF(__len) \ 247 (HH_DATA_MOD - (((__len - 1) & (HH_DATA_MOD - 1)) + 1)) 248 #define HH_DATA_ALIGN(__len) \ 249 (((__len)+(HH_DATA_MOD-1))&~(HH_DATA_MOD - 1)) 250 unsigned long hh_data[HH_DATA_ALIGN(LL_MAX_HEADER) / sizeof(long)]; 251 }; 252 253 /* Reserve HH_DATA_MOD byte-aligned hard_header_len, but at least that much. 254 * Alternative is: 255 * dev->hard_header_len ? (dev->hard_header_len + 256 * (HH_DATA_MOD - 1)) & ~(HH_DATA_MOD - 1) : 0 257 * 258 * We could use other alignment values, but we must maintain the 259 * relationship HH alignment <= LL alignment. 260 */ 261 #define LL_RESERVED_SPACE(dev) \ 262 ((((dev)->hard_header_len+(dev)->needed_headroom)&~(HH_DATA_MOD - 1)) + HH_DATA_MOD) 263 #define LL_RESERVED_SPACE_EXTRA(dev,extra) \ 264 ((((dev)->hard_header_len+(dev)->needed_headroom+(extra))&~(HH_DATA_MOD - 1)) + HH_DATA_MOD) 265 266 struct header_ops { 267 int (*create) (struct sk_buff *skb, struct net_device *dev, 268 unsigned short type, const void *daddr, 269 const void *saddr, unsigned int len); 270 int (*parse)(const struct sk_buff *skb, unsigned char *haddr); 271 int (*cache)(const struct neighbour *neigh, struct hh_cache *hh, __be16 type); 272 void (*cache_update)(struct hh_cache *hh, 273 const struct net_device *dev, 274 const unsigned char *haddr); 275 bool (*validate)(const char *ll_header, unsigned int len); 276 }; 277 278 /* These flag bits are private to the generic network queueing 279 * layer; they may not be explicitly referenced by any other 280 * code. 281 */ 282 283 enum netdev_state_t { 284 __LINK_STATE_START, 285 __LINK_STATE_PRESENT, 286 __LINK_STATE_NOCARRIER, 287 __LINK_STATE_LINKWATCH_PENDING, 288 __LINK_STATE_DORMANT, 289 }; 290 291 292 /* 293 * This structure holds boot-time configured netdevice settings. They 294 * are then used in the device probing. 295 */ 296 struct netdev_boot_setup { 297 char name[IFNAMSIZ]; 298 struct ifmap map; 299 }; 300 #define NETDEV_BOOT_SETUP_MAX 8 301 302 int __init netdev_boot_setup(char *str); 303 304 /* 305 * Structure for NAPI scheduling similar to tasklet but with weighting 306 */ 307 struct napi_struct { 308 /* The poll_list must only be managed by the entity which 309 * changes the state of the NAPI_STATE_SCHED bit. This means 310 * whoever atomically sets that bit can add this napi_struct 311 * to the per-CPU poll_list, and whoever clears that bit 312 * can remove from the list right before clearing the bit. 313 */ 314 struct list_head poll_list; 315 316 unsigned long state; 317 int weight; 318 unsigned int gro_count; 319 int (*poll)(struct napi_struct *, int); 320 #ifdef CONFIG_NETPOLL 321 int poll_owner; 322 #endif 323 struct net_device *dev; 324 struct sk_buff *gro_list; 325 struct sk_buff *skb; 326 struct hrtimer timer; 327 struct list_head dev_list; 328 struct hlist_node napi_hash_node; 329 unsigned int napi_id; 330 }; 331 332 enum { 333 NAPI_STATE_SCHED, /* Poll is scheduled */ 334 NAPI_STATE_MISSED, /* reschedule a napi */ 335 NAPI_STATE_DISABLE, /* Disable pending */ 336 NAPI_STATE_NPSVC, /* Netpoll - don't dequeue from poll_list */ 337 NAPI_STATE_HASHED, /* In NAPI hash (busy polling possible) */ 338 NAPI_STATE_NO_BUSY_POLL,/* Do not add in napi_hash, no busy polling */ 339 NAPI_STATE_IN_BUSY_POLL,/* sk_busy_loop() owns this NAPI */ 340 }; 341 342 enum { 343 NAPIF_STATE_SCHED = BIT(NAPI_STATE_SCHED), 344 NAPIF_STATE_MISSED = BIT(NAPI_STATE_MISSED), 345 NAPIF_STATE_DISABLE = BIT(NAPI_STATE_DISABLE), 346 NAPIF_STATE_NPSVC = BIT(NAPI_STATE_NPSVC), 347 NAPIF_STATE_HASHED = BIT(NAPI_STATE_HASHED), 348 NAPIF_STATE_NO_BUSY_POLL = BIT(NAPI_STATE_NO_BUSY_POLL), 349 NAPIF_STATE_IN_BUSY_POLL = BIT(NAPI_STATE_IN_BUSY_POLL), 350 }; 351 352 enum gro_result { 353 GRO_MERGED, 354 GRO_MERGED_FREE, 355 GRO_HELD, 356 GRO_NORMAL, 357 GRO_DROP, 358 GRO_CONSUMED, 359 }; 360 typedef enum gro_result gro_result_t; 361 362 /* 363 * enum rx_handler_result - Possible return values for rx_handlers. 364 * @RX_HANDLER_CONSUMED: skb was consumed by rx_handler, do not process it 365 * further. 366 * @RX_HANDLER_ANOTHER: Do another round in receive path. This is indicated in 367 * case skb->dev was changed by rx_handler. 368 * @RX_HANDLER_EXACT: Force exact delivery, no wildcard. 369 * @RX_HANDLER_PASS: Do nothing, pass the skb as if no rx_handler was called. 370 * 371 * rx_handlers are functions called from inside __netif_receive_skb(), to do 372 * special processing of the skb, prior to delivery to protocol handlers. 373 * 374 * Currently, a net_device can only have a single rx_handler registered. Trying 375 * to register a second rx_handler will return -EBUSY. 376 * 377 * To register a rx_handler on a net_device, use netdev_rx_handler_register(). 378 * To unregister a rx_handler on a net_device, use 379 * netdev_rx_handler_unregister(). 380 * 381 * Upon return, rx_handler is expected to tell __netif_receive_skb() what to 382 * do with the skb. 383 * 384 * If the rx_handler consumed the skb in some way, it should return 385 * RX_HANDLER_CONSUMED. This is appropriate when the rx_handler arranged for 386 * the skb to be delivered in some other way. 387 * 388 * If the rx_handler changed skb->dev, to divert the skb to another 389 * net_device, it should return RX_HANDLER_ANOTHER. The rx_handler for the 390 * new device will be called if it exists. 391 * 392 * If the rx_handler decides the skb should be ignored, it should return 393 * RX_HANDLER_EXACT. The skb will only be delivered to protocol handlers that 394 * are registered on exact device (ptype->dev == skb->dev). 395 * 396 * If the rx_handler didn't change skb->dev, but wants the skb to be normally 397 * delivered, it should return RX_HANDLER_PASS. 398 * 399 * A device without a registered rx_handler will behave as if rx_handler 400 * returned RX_HANDLER_PASS. 401 */ 402 403 enum rx_handler_result { 404 RX_HANDLER_CONSUMED, 405 RX_HANDLER_ANOTHER, 406 RX_HANDLER_EXACT, 407 RX_HANDLER_PASS, 408 }; 409 typedef enum rx_handler_result rx_handler_result_t; 410 typedef rx_handler_result_t rx_handler_func_t(struct sk_buff **pskb); 411 412 void __napi_schedule(struct napi_struct *n); 413 void __napi_schedule_irqoff(struct napi_struct *n); 414 415 static inline bool napi_disable_pending(struct napi_struct *n) 416 { 417 return test_bit(NAPI_STATE_DISABLE, &n->state); 418 } 419 420 bool napi_schedule_prep(struct napi_struct *n); 421 422 /** 423 * napi_schedule - schedule NAPI poll 424 * @n: NAPI context 425 * 426 * Schedule NAPI poll routine to be called if it is not already 427 * running. 428 */ 429 static inline void napi_schedule(struct napi_struct *n) 430 { 431 if (napi_schedule_prep(n)) 432 __napi_schedule(n); 433 } 434 435 /** 436 * napi_schedule_irqoff - schedule NAPI poll 437 * @n: NAPI context 438 * 439 * Variant of napi_schedule(), assuming hard irqs are masked. 440 */ 441 static inline void napi_schedule_irqoff(struct napi_struct *n) 442 { 443 if (napi_schedule_prep(n)) 444 __napi_schedule_irqoff(n); 445 } 446 447 /* Try to reschedule poll. Called by dev->poll() after napi_complete(). */ 448 static inline bool napi_reschedule(struct napi_struct *napi) 449 { 450 if (napi_schedule_prep(napi)) { 451 __napi_schedule(napi); 452 return true; 453 } 454 return false; 455 } 456 457 bool napi_complete_done(struct napi_struct *n, int work_done); 458 /** 459 * napi_complete - NAPI processing complete 460 * @n: NAPI context 461 * 462 * Mark NAPI processing as complete. 463 * Consider using napi_complete_done() instead. 464 * Return false if device should avoid rearming interrupts. 465 */ 466 static inline bool napi_complete(struct napi_struct *n) 467 { 468 return napi_complete_done(n, 0); 469 } 470 471 /** 472 * napi_hash_del - remove a NAPI from global table 473 * @napi: NAPI context 474 * 475 * Warning: caller must observe RCU grace period 476 * before freeing memory containing @napi, if 477 * this function returns true. 478 * Note: core networking stack automatically calls it 479 * from netif_napi_del(). 480 * Drivers might want to call this helper to combine all 481 * the needed RCU grace periods into a single one. 482 */ 483 bool napi_hash_del(struct napi_struct *napi); 484 485 /** 486 * napi_disable - prevent NAPI from scheduling 487 * @n: NAPI context 488 * 489 * Stop NAPI from being scheduled on this context. 490 * Waits till any outstanding processing completes. 491 */ 492 void napi_disable(struct napi_struct *n); 493 494 /** 495 * napi_enable - enable NAPI scheduling 496 * @n: NAPI context 497 * 498 * Resume NAPI from being scheduled on this context. 499 * Must be paired with napi_disable. 500 */ 501 static inline void napi_enable(struct napi_struct *n) 502 { 503 BUG_ON(!test_bit(NAPI_STATE_SCHED, &n->state)); 504 smp_mb__before_atomic(); 505 clear_bit(NAPI_STATE_SCHED, &n->state); 506 clear_bit(NAPI_STATE_NPSVC, &n->state); 507 } 508 509 /** 510 * napi_synchronize - wait until NAPI is not running 511 * @n: NAPI context 512 * 513 * Wait until NAPI is done being scheduled on this context. 514 * Waits till any outstanding processing completes but 515 * does not disable future activations. 516 */ 517 static inline void napi_synchronize(const struct napi_struct *n) 518 { 519 if (IS_ENABLED(CONFIG_SMP)) 520 while (test_bit(NAPI_STATE_SCHED, &n->state)) 521 msleep(1); 522 else 523 barrier(); 524 } 525 526 enum netdev_queue_state_t { 527 __QUEUE_STATE_DRV_XOFF, 528 __QUEUE_STATE_STACK_XOFF, 529 __QUEUE_STATE_FROZEN, 530 }; 531 532 #define QUEUE_STATE_DRV_XOFF (1 << __QUEUE_STATE_DRV_XOFF) 533 #define QUEUE_STATE_STACK_XOFF (1 << __QUEUE_STATE_STACK_XOFF) 534 #define QUEUE_STATE_FROZEN (1 << __QUEUE_STATE_FROZEN) 535 536 #define QUEUE_STATE_ANY_XOFF (QUEUE_STATE_DRV_XOFF | QUEUE_STATE_STACK_XOFF) 537 #define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \ 538 QUEUE_STATE_FROZEN) 539 #define QUEUE_STATE_DRV_XOFF_OR_FROZEN (QUEUE_STATE_DRV_XOFF | \ 540 QUEUE_STATE_FROZEN) 541 542 /* 543 * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue. The 544 * netif_tx_* functions below are used to manipulate this flag. The 545 * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit 546 * queue independently. The netif_xmit_*stopped functions below are called 547 * to check if the queue has been stopped by the driver or stack (either 548 * of the XOFF bits are set in the state). Drivers should not need to call 549 * netif_xmit*stopped functions, they should only be using netif_tx_*. 550 */ 551 552 struct netdev_queue { 553 /* 554 * read-mostly part 555 */ 556 struct net_device *dev; 557 struct Qdisc __rcu *qdisc; 558 struct Qdisc *qdisc_sleeping; 559 #ifdef CONFIG_SYSFS 560 struct kobject kobj; 561 #endif 562 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA) 563 int numa_node; 564 #endif 565 unsigned long tx_maxrate; 566 /* 567 * Number of TX timeouts for this queue 568 * (/sys/class/net/DEV/Q/trans_timeout) 569 */ 570 unsigned long trans_timeout; 571 /* 572 * write-mostly part 573 */ 574 spinlock_t _xmit_lock ____cacheline_aligned_in_smp; 575 int xmit_lock_owner; 576 /* 577 * Time (in jiffies) of last Tx 578 */ 579 unsigned long trans_start; 580 581 unsigned long state; 582 583 #ifdef CONFIG_BQL 584 struct dql dql; 585 #endif 586 } ____cacheline_aligned_in_smp; 587 588 static inline int netdev_queue_numa_node_read(const struct netdev_queue *q) 589 { 590 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA) 591 return q->numa_node; 592 #else 593 return NUMA_NO_NODE; 594 #endif 595 } 596 597 static inline void netdev_queue_numa_node_write(struct netdev_queue *q, int node) 598 { 599 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA) 600 q->numa_node = node; 601 #endif 602 } 603 604 #ifdef CONFIG_RPS 605 /* 606 * This structure holds an RPS map which can be of variable length. The 607 * map is an array of CPUs. 608 */ 609 struct rps_map { 610 unsigned int len; 611 struct rcu_head rcu; 612 u16 cpus[0]; 613 }; 614 #define RPS_MAP_SIZE(_num) (sizeof(struct rps_map) + ((_num) * sizeof(u16))) 615 616 /* 617 * The rps_dev_flow structure contains the mapping of a flow to a CPU, the 618 * tail pointer for that CPU's input queue at the time of last enqueue, and 619 * a hardware filter index. 620 */ 621 struct rps_dev_flow { 622 u16 cpu; 623 u16 filter; 624 unsigned int last_qtail; 625 }; 626 #define RPS_NO_FILTER 0xffff 627 628 /* 629 * The rps_dev_flow_table structure contains a table of flow mappings. 630 */ 631 struct rps_dev_flow_table { 632 unsigned int mask; 633 struct rcu_head rcu; 634 struct rps_dev_flow flows[0]; 635 }; 636 #define RPS_DEV_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_dev_flow_table) + \ 637 ((_num) * sizeof(struct rps_dev_flow))) 638 639 /* 640 * The rps_sock_flow_table contains mappings of flows to the last CPU 641 * on which they were processed by the application (set in recvmsg). 642 * Each entry is a 32bit value. Upper part is the high-order bits 643 * of flow hash, lower part is CPU number. 644 * rps_cpu_mask is used to partition the space, depending on number of 645 * possible CPUs : rps_cpu_mask = roundup_pow_of_two(nr_cpu_ids) - 1 646 * For example, if 64 CPUs are possible, rps_cpu_mask = 0x3f, 647 * meaning we use 32-6=26 bits for the hash. 648 */ 649 struct rps_sock_flow_table { 650 u32 mask; 651 652 u32 ents[0] ____cacheline_aligned_in_smp; 653 }; 654 #define RPS_SOCK_FLOW_TABLE_SIZE(_num) (offsetof(struct rps_sock_flow_table, ents[_num])) 655 656 #define RPS_NO_CPU 0xffff 657 658 extern u32 rps_cpu_mask; 659 extern struct rps_sock_flow_table __rcu *rps_sock_flow_table; 660 661 static inline void rps_record_sock_flow(struct rps_sock_flow_table *table, 662 u32 hash) 663 { 664 if (table && hash) { 665 unsigned int index = hash & table->mask; 666 u32 val = hash & ~rps_cpu_mask; 667 668 /* We only give a hint, preemption can change CPU under us */ 669 val |= raw_smp_processor_id(); 670 671 if (table->ents[index] != val) 672 table->ents[index] = val; 673 } 674 } 675 676 #ifdef CONFIG_RFS_ACCEL 677 bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index, u32 flow_id, 678 u16 filter_id); 679 #endif 680 #endif /* CONFIG_RPS */ 681 682 /* This structure contains an instance of an RX queue. */ 683 struct netdev_rx_queue { 684 #ifdef CONFIG_RPS 685 struct rps_map __rcu *rps_map; 686 struct rps_dev_flow_table __rcu *rps_flow_table; 687 #endif 688 struct kobject kobj; 689 struct net_device *dev; 690 } ____cacheline_aligned_in_smp; 691 692 /* 693 * RX queue sysfs structures and functions. 694 */ 695 struct rx_queue_attribute { 696 struct attribute attr; 697 ssize_t (*show)(struct netdev_rx_queue *queue, char *buf); 698 ssize_t (*store)(struct netdev_rx_queue *queue, 699 const char *buf, size_t len); 700 }; 701 702 #ifdef CONFIG_XPS 703 /* 704 * This structure holds an XPS map which can be of variable length. The 705 * map is an array of queues. 706 */ 707 struct xps_map { 708 unsigned int len; 709 unsigned int alloc_len; 710 struct rcu_head rcu; 711 u16 queues[0]; 712 }; 713 #define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16))) 714 #define XPS_MIN_MAP_ALLOC ((L1_CACHE_ALIGN(offsetof(struct xps_map, queues[1])) \ 715 - sizeof(struct xps_map)) / sizeof(u16)) 716 717 /* 718 * This structure holds all XPS maps for device. Maps are indexed by CPU. 719 */ 720 struct xps_dev_maps { 721 struct rcu_head rcu; 722 struct xps_map __rcu *cpu_map[0]; 723 }; 724 #define XPS_DEV_MAPS_SIZE(_tcs) (sizeof(struct xps_dev_maps) + \ 725 (nr_cpu_ids * (_tcs) * sizeof(struct xps_map *))) 726 #endif /* CONFIG_XPS */ 727 728 #define TC_MAX_QUEUE 16 729 #define TC_BITMASK 15 730 /* HW offloaded queuing disciplines txq count and offset maps */ 731 struct netdev_tc_txq { 732 u16 count; 733 u16 offset; 734 }; 735 736 #if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE) 737 /* 738 * This structure is to hold information about the device 739 * configured to run FCoE protocol stack. 740 */ 741 struct netdev_fcoe_hbainfo { 742 char manufacturer[64]; 743 char serial_number[64]; 744 char hardware_version[64]; 745 char driver_version[64]; 746 char optionrom_version[64]; 747 char firmware_version[64]; 748 char model[256]; 749 char model_description[256]; 750 }; 751 #endif 752 753 #define MAX_PHYS_ITEM_ID_LEN 32 754 755 /* This structure holds a unique identifier to identify some 756 * physical item (port for example) used by a netdevice. 757 */ 758 struct netdev_phys_item_id { 759 unsigned char id[MAX_PHYS_ITEM_ID_LEN]; 760 unsigned char id_len; 761 }; 762 763 static inline bool netdev_phys_item_id_same(struct netdev_phys_item_id *a, 764 struct netdev_phys_item_id *b) 765 { 766 return a->id_len == b->id_len && 767 memcmp(a->id, b->id, a->id_len) == 0; 768 } 769 770 typedef u16 (*select_queue_fallback_t)(struct net_device *dev, 771 struct sk_buff *skb); 772 773 enum tc_setup_type { 774 TC_SETUP_MQPRIO, 775 TC_SETUP_CLSU32, 776 TC_SETUP_CLSFLOWER, 777 TC_SETUP_CLSMATCHALL, 778 TC_SETUP_CLSBPF, 779 }; 780 781 /* These structures hold the attributes of xdp state that are being passed 782 * to the netdevice through the xdp op. 783 */ 784 enum xdp_netdev_command { 785 /* Set or clear a bpf program used in the earliest stages of packet 786 * rx. The prog will have been loaded as BPF_PROG_TYPE_XDP. The callee 787 * is responsible for calling bpf_prog_put on any old progs that are 788 * stored. In case of error, the callee need not release the new prog 789 * reference, but on success it takes ownership and must bpf_prog_put 790 * when it is no longer used. 791 */ 792 XDP_SETUP_PROG, 793 XDP_SETUP_PROG_HW, 794 /* Check if a bpf program is set on the device. The callee should 795 * set @prog_attached to one of XDP_ATTACHED_* values, note that "true" 796 * is equivalent to XDP_ATTACHED_DRV. 797 */ 798 XDP_QUERY_PROG, 799 }; 800 801 struct netlink_ext_ack; 802 803 struct netdev_xdp { 804 enum xdp_netdev_command command; 805 union { 806 /* XDP_SETUP_PROG */ 807 struct { 808 u32 flags; 809 struct bpf_prog *prog; 810 struct netlink_ext_ack *extack; 811 }; 812 /* XDP_QUERY_PROG */ 813 struct { 814 u8 prog_attached; 815 u32 prog_id; 816 }; 817 }; 818 }; 819 820 #ifdef CONFIG_XFRM_OFFLOAD 821 struct xfrmdev_ops { 822 int (*xdo_dev_state_add) (struct xfrm_state *x); 823 void (*xdo_dev_state_delete) (struct xfrm_state *x); 824 void (*xdo_dev_state_free) (struct xfrm_state *x); 825 bool (*xdo_dev_offload_ok) (struct sk_buff *skb, 826 struct xfrm_state *x); 827 }; 828 #endif 829 830 /* 831 * This structure defines the management hooks for network devices. 832 * The following hooks can be defined; unless noted otherwise, they are 833 * optional and can be filled with a null pointer. 834 * 835 * int (*ndo_init)(struct net_device *dev); 836 * This function is called once when a network device is registered. 837 * The network device can use this for any late stage initialization 838 * or semantic validation. It can fail with an error code which will 839 * be propagated back to register_netdev. 840 * 841 * void (*ndo_uninit)(struct net_device *dev); 842 * This function is called when device is unregistered or when registration 843 * fails. It is not called if init fails. 844 * 845 * int (*ndo_open)(struct net_device *dev); 846 * This function is called when a network device transitions to the up 847 * state. 848 * 849 * int (*ndo_stop)(struct net_device *dev); 850 * This function is called when a network device transitions to the down 851 * state. 852 * 853 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb, 854 * struct net_device *dev); 855 * Called when a packet needs to be transmitted. 856 * Returns NETDEV_TX_OK. Can return NETDEV_TX_BUSY, but you should stop 857 * the queue before that can happen; it's for obsolete devices and weird 858 * corner cases, but the stack really does a non-trivial amount 859 * of useless work if you return NETDEV_TX_BUSY. 860 * Required; cannot be NULL. 861 * 862 * netdev_features_t (*ndo_features_check)(struct sk_buff *skb, 863 * struct net_device *dev 864 * netdev_features_t features); 865 * Called by core transmit path to determine if device is capable of 866 * performing offload operations on a given packet. This is to give 867 * the device an opportunity to implement any restrictions that cannot 868 * be otherwise expressed by feature flags. The check is called with 869 * the set of features that the stack has calculated and it returns 870 * those the driver believes to be appropriate. 871 * 872 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb, 873 * void *accel_priv, select_queue_fallback_t fallback); 874 * Called to decide which queue to use when device supports multiple 875 * transmit queues. 876 * 877 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags); 878 * This function is called to allow device receiver to make 879 * changes to configuration when multicast or promiscuous is enabled. 880 * 881 * void (*ndo_set_rx_mode)(struct net_device *dev); 882 * This function is called device changes address list filtering. 883 * If driver handles unicast address filtering, it should set 884 * IFF_UNICAST_FLT in its priv_flags. 885 * 886 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr); 887 * This function is called when the Media Access Control address 888 * needs to be changed. If this interface is not defined, the 889 * MAC address can not be changed. 890 * 891 * int (*ndo_validate_addr)(struct net_device *dev); 892 * Test if Media Access Control address is valid for the device. 893 * 894 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd); 895 * Called when a user requests an ioctl which can't be handled by 896 * the generic interface code. If not defined ioctls return 897 * not supported error code. 898 * 899 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map); 900 * Used to set network devices bus interface parameters. This interface 901 * is retained for legacy reasons; new devices should use the bus 902 * interface (PCI) for low level management. 903 * 904 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu); 905 * Called when a user wants to change the Maximum Transfer Unit 906 * of a device. 907 * 908 * void (*ndo_tx_timeout)(struct net_device *dev); 909 * Callback used when the transmitter has not made any progress 910 * for dev->watchdog ticks. 911 * 912 * void (*ndo_get_stats64)(struct net_device *dev, 913 * struct rtnl_link_stats64 *storage); 914 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev); 915 * Called when a user wants to get the network device usage 916 * statistics. Drivers must do one of the following: 917 * 1. Define @ndo_get_stats64 to fill in a zero-initialised 918 * rtnl_link_stats64 structure passed by the caller. 919 * 2. Define @ndo_get_stats to update a net_device_stats structure 920 * (which should normally be dev->stats) and return a pointer to 921 * it. The structure may be changed asynchronously only if each 922 * field is written atomically. 923 * 3. Update dev->stats asynchronously and atomically, and define 924 * neither operation. 925 * 926 * bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id) 927 * Return true if this device supports offload stats of this attr_id. 928 * 929 * int (*ndo_get_offload_stats)(int attr_id, const struct net_device *dev, 930 * void *attr_data) 931 * Get statistics for offload operations by attr_id. Write it into the 932 * attr_data pointer. 933 * 934 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16 vid); 935 * If device supports VLAN filtering this function is called when a 936 * VLAN id is registered. 937 * 938 * int (*ndo_vlan_rx_kill_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 unregistered. 941 * 942 * void (*ndo_poll_controller)(struct net_device *dev); 943 * 944 * SR-IOV management functions. 945 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac); 946 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan, 947 * u8 qos, __be16 proto); 948 * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate, 949 * int max_tx_rate); 950 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting); 951 * int (*ndo_set_vf_trust)(struct net_device *dev, int vf, bool setting); 952 * int (*ndo_get_vf_config)(struct net_device *dev, 953 * int vf, struct ifla_vf_info *ivf); 954 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state); 955 * int (*ndo_set_vf_port)(struct net_device *dev, int vf, 956 * struct nlattr *port[]); 957 * 958 * Enable or disable the VF ability to query its RSS Redirection Table and 959 * Hash Key. This is needed since on some devices VF share this information 960 * with PF and querying it may introduce a theoretical security risk. 961 * int (*ndo_set_vf_rss_query_en)(struct net_device *dev, int vf, bool setting); 962 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb); 963 * int (*ndo_setup_tc)(struct net_device *dev, enum tc_setup_type type, 964 * void *type_data); 965 * Called to setup any 'tc' scheduler, classifier or action on @dev. 966 * This is always called from the stack with the rtnl lock held and netif 967 * tx queues stopped. This allows the netdevice to perform queue 968 * management safely. 969 * 970 * Fiber Channel over Ethernet (FCoE) offload functions. 971 * int (*ndo_fcoe_enable)(struct net_device *dev); 972 * Called when the FCoE protocol stack wants to start using LLD for FCoE 973 * so the underlying device can perform whatever needed configuration or 974 * initialization to support acceleration of FCoE traffic. 975 * 976 * int (*ndo_fcoe_disable)(struct net_device *dev); 977 * Called when the FCoE protocol stack wants to stop using LLD for FCoE 978 * so the underlying device can perform whatever needed clean-ups to 979 * stop supporting acceleration of FCoE traffic. 980 * 981 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid, 982 * struct scatterlist *sgl, unsigned int sgc); 983 * Called when the FCoE Initiator wants to initialize an I/O that 984 * is a possible candidate for Direct Data Placement (DDP). The LLD can 985 * perform necessary setup and returns 1 to indicate the device is set up 986 * successfully to perform DDP on this I/O, otherwise this returns 0. 987 * 988 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid); 989 * Called when the FCoE Initiator/Target is done with the DDPed I/O as 990 * indicated by the FC exchange id 'xid', so the underlying device can 991 * clean up and reuse resources for later DDP requests. 992 * 993 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid, 994 * struct scatterlist *sgl, unsigned int sgc); 995 * Called when the FCoE Target wants to initialize an I/O that 996 * is a possible candidate for Direct Data Placement (DDP). The LLD can 997 * perform necessary setup and returns 1 to indicate the device is set up 998 * successfully to perform DDP on this I/O, otherwise this returns 0. 999 * 1000 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev, 1001 * struct netdev_fcoe_hbainfo *hbainfo); 1002 * Called when the FCoE Protocol stack wants information on the underlying 1003 * device. This information is utilized by the FCoE protocol stack to 1004 * register attributes with Fiber Channel management service as per the 1005 * FC-GS Fabric Device Management Information(FDMI) specification. 1006 * 1007 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type); 1008 * Called when the underlying device wants to override default World Wide 1009 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own 1010 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE 1011 * protocol stack to use. 1012 * 1013 * RFS acceleration. 1014 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb, 1015 * u16 rxq_index, u32 flow_id); 1016 * Set hardware filter for RFS. rxq_index is the target queue index; 1017 * flow_id is a flow ID to be passed to rps_may_expire_flow() later. 1018 * Return the filter ID on success, or a negative error code. 1019 * 1020 * Slave management functions (for bridge, bonding, etc). 1021 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev); 1022 * Called to make another netdev an underling. 1023 * 1024 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev); 1025 * Called to release previously enslaved netdev. 1026 * 1027 * Feature/offload setting functions. 1028 * netdev_features_t (*ndo_fix_features)(struct net_device *dev, 1029 * netdev_features_t features); 1030 * Adjusts the requested feature flags according to device-specific 1031 * constraints, and returns the resulting flags. Must not modify 1032 * the device state. 1033 * 1034 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features); 1035 * Called to update device configuration to new features. Passed 1036 * feature set might be less than what was returned by ndo_fix_features()). 1037 * Must return >0 or -errno if it changed dev->features itself. 1038 * 1039 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[], 1040 * struct net_device *dev, 1041 * const unsigned char *addr, u16 vid, u16 flags) 1042 * Adds an FDB entry to dev for addr. 1043 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[], 1044 * struct net_device *dev, 1045 * const unsigned char *addr, u16 vid) 1046 * Deletes the FDB entry from dev coresponding to addr. 1047 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb, 1048 * struct net_device *dev, struct net_device *filter_dev, 1049 * int *idx) 1050 * Used to add FDB entries to dump requests. Implementers should add 1051 * entries to skb and update idx with the number of entries. 1052 * 1053 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh, 1054 * u16 flags) 1055 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq, 1056 * struct net_device *dev, u32 filter_mask, 1057 * int nlflags) 1058 * int (*ndo_bridge_dellink)(struct net_device *dev, struct nlmsghdr *nlh, 1059 * u16 flags); 1060 * 1061 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier); 1062 * Called to change device carrier. Soft-devices (like dummy, team, etc) 1063 * which do not represent real hardware may define this to allow their 1064 * userspace components to manage their virtual carrier state. Devices 1065 * that determine carrier state from physical hardware properties (eg 1066 * network cables) or protocol-dependent mechanisms (eg 1067 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function. 1068 * 1069 * int (*ndo_get_phys_port_id)(struct net_device *dev, 1070 * struct netdev_phys_item_id *ppid); 1071 * Called to get ID of physical port of this device. If driver does 1072 * not implement this, it is assumed that the hw is not able to have 1073 * multiple net devices on single physical port. 1074 * 1075 * void (*ndo_udp_tunnel_add)(struct net_device *dev, 1076 * struct udp_tunnel_info *ti); 1077 * Called by UDP tunnel to notify a driver about the UDP port and socket 1078 * address family that a UDP tunnel is listnening to. It is called only 1079 * when a new port starts listening. The operation is protected by the 1080 * RTNL. 1081 * 1082 * void (*ndo_udp_tunnel_del)(struct net_device *dev, 1083 * struct udp_tunnel_info *ti); 1084 * Called by UDP tunnel to notify the driver about a UDP port and socket 1085 * address family that the UDP tunnel is not listening to anymore. The 1086 * operation is protected by the RTNL. 1087 * 1088 * void* (*ndo_dfwd_add_station)(struct net_device *pdev, 1089 * struct net_device *dev) 1090 * Called by upper layer devices to accelerate switching or other 1091 * station functionality into hardware. 'pdev is the lowerdev 1092 * to use for the offload and 'dev' is the net device that will 1093 * back the offload. Returns a pointer to the private structure 1094 * the upper layer will maintain. 1095 * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv) 1096 * Called by upper layer device to delete the station created 1097 * by 'ndo_dfwd_add_station'. 'pdev' is the net device backing 1098 * the station and priv is the structure returned by the add 1099 * operation. 1100 * int (*ndo_set_tx_maxrate)(struct net_device *dev, 1101 * int queue_index, u32 maxrate); 1102 * Called when a user wants to set a max-rate limitation of specific 1103 * TX queue. 1104 * int (*ndo_get_iflink)(const struct net_device *dev); 1105 * Called to get the iflink value of this device. 1106 * void (*ndo_change_proto_down)(struct net_device *dev, 1107 * bool proto_down); 1108 * This function is used to pass protocol port error state information 1109 * to the switch driver. The switch driver can react to the proto_down 1110 * by doing a phys down on the associated switch port. 1111 * int (*ndo_fill_metadata_dst)(struct net_device *dev, struct sk_buff *skb); 1112 * This function is used to get egress tunnel information for given skb. 1113 * This is useful for retrieving outer tunnel header parameters while 1114 * sampling packet. 1115 * void (*ndo_set_rx_headroom)(struct net_device *dev, int needed_headroom); 1116 * This function is used to specify the headroom that the skb must 1117 * consider when allocation skb during packet reception. Setting 1118 * appropriate rx headroom value allows avoiding skb head copy on 1119 * forward. Setting a negative value resets the rx headroom to the 1120 * default value. 1121 * int (*ndo_xdp)(struct net_device *dev, struct netdev_xdp *xdp); 1122 * This function is used to set or query state related to XDP on the 1123 * netdevice. See definition of enum xdp_netdev_command for details. 1124 * int (*ndo_xdp_xmit)(struct net_device *dev, struct xdp_buff *xdp); 1125 * This function is used to submit a XDP packet for transmit on a 1126 * netdevice. 1127 * void (*ndo_xdp_flush)(struct net_device *dev); 1128 * This function is used to inform the driver to flush a paticular 1129 * xpd tx queue. Must be called on same CPU as xdp_xmit. 1130 */ 1131 struct net_device_ops { 1132 int (*ndo_init)(struct net_device *dev); 1133 void (*ndo_uninit)(struct net_device *dev); 1134 int (*ndo_open)(struct net_device *dev); 1135 int (*ndo_stop)(struct net_device *dev); 1136 netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb, 1137 struct net_device *dev); 1138 netdev_features_t (*ndo_features_check)(struct sk_buff *skb, 1139 struct net_device *dev, 1140 netdev_features_t features); 1141 u16 (*ndo_select_queue)(struct net_device *dev, 1142 struct sk_buff *skb, 1143 void *accel_priv, 1144 select_queue_fallback_t fallback); 1145 void (*ndo_change_rx_flags)(struct net_device *dev, 1146 int flags); 1147 void (*ndo_set_rx_mode)(struct net_device *dev); 1148 int (*ndo_set_mac_address)(struct net_device *dev, 1149 void *addr); 1150 int (*ndo_validate_addr)(struct net_device *dev); 1151 int (*ndo_do_ioctl)(struct net_device *dev, 1152 struct ifreq *ifr, int cmd); 1153 int (*ndo_set_config)(struct net_device *dev, 1154 struct ifmap *map); 1155 int (*ndo_change_mtu)(struct net_device *dev, 1156 int new_mtu); 1157 int (*ndo_neigh_setup)(struct net_device *dev, 1158 struct neigh_parms *); 1159 void (*ndo_tx_timeout) (struct net_device *dev); 1160 1161 void (*ndo_get_stats64)(struct net_device *dev, 1162 struct rtnl_link_stats64 *storage); 1163 bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id); 1164 int (*ndo_get_offload_stats)(int attr_id, 1165 const struct net_device *dev, 1166 void *attr_data); 1167 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev); 1168 1169 int (*ndo_vlan_rx_add_vid)(struct net_device *dev, 1170 __be16 proto, u16 vid); 1171 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, 1172 __be16 proto, u16 vid); 1173 #ifdef CONFIG_NET_POLL_CONTROLLER 1174 void (*ndo_poll_controller)(struct net_device *dev); 1175 int (*ndo_netpoll_setup)(struct net_device *dev, 1176 struct netpoll_info *info); 1177 void (*ndo_netpoll_cleanup)(struct net_device *dev); 1178 #endif 1179 int (*ndo_set_vf_mac)(struct net_device *dev, 1180 int queue, u8 *mac); 1181 int (*ndo_set_vf_vlan)(struct net_device *dev, 1182 int queue, u16 vlan, 1183 u8 qos, __be16 proto); 1184 int (*ndo_set_vf_rate)(struct net_device *dev, 1185 int vf, int min_tx_rate, 1186 int max_tx_rate); 1187 int (*ndo_set_vf_spoofchk)(struct net_device *dev, 1188 int vf, bool setting); 1189 int (*ndo_set_vf_trust)(struct net_device *dev, 1190 int vf, bool setting); 1191 int (*ndo_get_vf_config)(struct net_device *dev, 1192 int vf, 1193 struct ifla_vf_info *ivf); 1194 int (*ndo_set_vf_link_state)(struct net_device *dev, 1195 int vf, int link_state); 1196 int (*ndo_get_vf_stats)(struct net_device *dev, 1197 int vf, 1198 struct ifla_vf_stats 1199 *vf_stats); 1200 int (*ndo_set_vf_port)(struct net_device *dev, 1201 int vf, 1202 struct nlattr *port[]); 1203 int (*ndo_get_vf_port)(struct net_device *dev, 1204 int vf, struct sk_buff *skb); 1205 int (*ndo_set_vf_guid)(struct net_device *dev, 1206 int vf, u64 guid, 1207 int guid_type); 1208 int (*ndo_set_vf_rss_query_en)( 1209 struct net_device *dev, 1210 int vf, bool setting); 1211 int (*ndo_setup_tc)(struct net_device *dev, 1212 enum tc_setup_type type, 1213 void *type_data); 1214 #if IS_ENABLED(CONFIG_FCOE) 1215 int (*ndo_fcoe_enable)(struct net_device *dev); 1216 int (*ndo_fcoe_disable)(struct net_device *dev); 1217 int (*ndo_fcoe_ddp_setup)(struct net_device *dev, 1218 u16 xid, 1219 struct scatterlist *sgl, 1220 unsigned int sgc); 1221 int (*ndo_fcoe_ddp_done)(struct net_device *dev, 1222 u16 xid); 1223 int (*ndo_fcoe_ddp_target)(struct net_device *dev, 1224 u16 xid, 1225 struct scatterlist *sgl, 1226 unsigned int sgc); 1227 int (*ndo_fcoe_get_hbainfo)(struct net_device *dev, 1228 struct netdev_fcoe_hbainfo *hbainfo); 1229 #endif 1230 1231 #if IS_ENABLED(CONFIG_LIBFCOE) 1232 #define NETDEV_FCOE_WWNN 0 1233 #define NETDEV_FCOE_WWPN 1 1234 int (*ndo_fcoe_get_wwn)(struct net_device *dev, 1235 u64 *wwn, int type); 1236 #endif 1237 1238 #ifdef CONFIG_RFS_ACCEL 1239 int (*ndo_rx_flow_steer)(struct net_device *dev, 1240 const struct sk_buff *skb, 1241 u16 rxq_index, 1242 u32 flow_id); 1243 #endif 1244 int (*ndo_add_slave)(struct net_device *dev, 1245 struct net_device *slave_dev); 1246 int (*ndo_del_slave)(struct net_device *dev, 1247 struct net_device *slave_dev); 1248 netdev_features_t (*ndo_fix_features)(struct net_device *dev, 1249 netdev_features_t features); 1250 int (*ndo_set_features)(struct net_device *dev, 1251 netdev_features_t features); 1252 int (*ndo_neigh_construct)(struct net_device *dev, 1253 struct neighbour *n); 1254 void (*ndo_neigh_destroy)(struct net_device *dev, 1255 struct neighbour *n); 1256 1257 int (*ndo_fdb_add)(struct ndmsg *ndm, 1258 struct nlattr *tb[], 1259 struct net_device *dev, 1260 const unsigned char *addr, 1261 u16 vid, 1262 u16 flags); 1263 int (*ndo_fdb_del)(struct ndmsg *ndm, 1264 struct nlattr *tb[], 1265 struct net_device *dev, 1266 const unsigned char *addr, 1267 u16 vid); 1268 int (*ndo_fdb_dump)(struct sk_buff *skb, 1269 struct netlink_callback *cb, 1270 struct net_device *dev, 1271 struct net_device *filter_dev, 1272 int *idx); 1273 1274 int (*ndo_bridge_setlink)(struct net_device *dev, 1275 struct nlmsghdr *nlh, 1276 u16 flags); 1277 int (*ndo_bridge_getlink)(struct sk_buff *skb, 1278 u32 pid, u32 seq, 1279 struct net_device *dev, 1280 u32 filter_mask, 1281 int nlflags); 1282 int (*ndo_bridge_dellink)(struct net_device *dev, 1283 struct nlmsghdr *nlh, 1284 u16 flags); 1285 int (*ndo_change_carrier)(struct net_device *dev, 1286 bool new_carrier); 1287 int (*ndo_get_phys_port_id)(struct net_device *dev, 1288 struct netdev_phys_item_id *ppid); 1289 int (*ndo_get_phys_port_name)(struct net_device *dev, 1290 char *name, size_t len); 1291 void (*ndo_udp_tunnel_add)(struct net_device *dev, 1292 struct udp_tunnel_info *ti); 1293 void (*ndo_udp_tunnel_del)(struct net_device *dev, 1294 struct udp_tunnel_info *ti); 1295 void* (*ndo_dfwd_add_station)(struct net_device *pdev, 1296 struct net_device *dev); 1297 void (*ndo_dfwd_del_station)(struct net_device *pdev, 1298 void *priv); 1299 1300 int (*ndo_get_lock_subclass)(struct net_device *dev); 1301 int (*ndo_set_tx_maxrate)(struct net_device *dev, 1302 int queue_index, 1303 u32 maxrate); 1304 int (*ndo_get_iflink)(const struct net_device *dev); 1305 int (*ndo_change_proto_down)(struct net_device *dev, 1306 bool proto_down); 1307 int (*ndo_fill_metadata_dst)(struct net_device *dev, 1308 struct sk_buff *skb); 1309 void (*ndo_set_rx_headroom)(struct net_device *dev, 1310 int needed_headroom); 1311 int (*ndo_xdp)(struct net_device *dev, 1312 struct netdev_xdp *xdp); 1313 int (*ndo_xdp_xmit)(struct net_device *dev, 1314 struct xdp_buff *xdp); 1315 void (*ndo_xdp_flush)(struct net_device *dev); 1316 }; 1317 1318 /** 1319 * enum net_device_priv_flags - &struct net_device priv_flags 1320 * 1321 * These are the &struct net_device, they are only set internally 1322 * by drivers and used in the kernel. These flags are invisible to 1323 * userspace; this means that the order of these flags can change 1324 * during any kernel release. 1325 * 1326 * You should have a pretty good reason to be extending these flags. 1327 * 1328 * @IFF_802_1Q_VLAN: 802.1Q VLAN device 1329 * @IFF_EBRIDGE: Ethernet bridging device 1330 * @IFF_BONDING: bonding master or slave 1331 * @IFF_ISATAP: ISATAP interface (RFC4214) 1332 * @IFF_WAN_HDLC: WAN HDLC device 1333 * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to 1334 * release skb->dst 1335 * @IFF_DONT_BRIDGE: disallow bridging this ether dev 1336 * @IFF_DISABLE_NETPOLL: disable netpoll at run-time 1337 * @IFF_MACVLAN_PORT: device used as macvlan port 1338 * @IFF_BRIDGE_PORT: device used as bridge port 1339 * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port 1340 * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit 1341 * @IFF_UNICAST_FLT: Supports unicast filtering 1342 * @IFF_TEAM_PORT: device used as team port 1343 * @IFF_SUPP_NOFCS: device supports sending custom FCS 1344 * @IFF_LIVE_ADDR_CHANGE: device supports hardware address 1345 * change when it's running 1346 * @IFF_MACVLAN: Macvlan device 1347 * @IFF_XMIT_DST_RELEASE_PERM: IFF_XMIT_DST_RELEASE not taking into account 1348 * underlying stacked devices 1349 * @IFF_IPVLAN_MASTER: IPvlan master device 1350 * @IFF_IPVLAN_SLAVE: IPvlan slave device 1351 * @IFF_L3MDEV_MASTER: device is an L3 master device 1352 * @IFF_NO_QUEUE: device can run without qdisc attached 1353 * @IFF_OPENVSWITCH: device is a Open vSwitch master 1354 * @IFF_L3MDEV_SLAVE: device is enslaved to an L3 master device 1355 * @IFF_TEAM: device is a team device 1356 * @IFF_RXFH_CONFIGURED: device has had Rx Flow indirection table configured 1357 * @IFF_PHONY_HEADROOM: the headroom value is controlled by an external 1358 * entity (i.e. the master device for bridged veth) 1359 * @IFF_MACSEC: device is a MACsec device 1360 */ 1361 enum netdev_priv_flags { 1362 IFF_802_1Q_VLAN = 1<<0, 1363 IFF_EBRIDGE = 1<<1, 1364 IFF_BONDING = 1<<2, 1365 IFF_ISATAP = 1<<3, 1366 IFF_WAN_HDLC = 1<<4, 1367 IFF_XMIT_DST_RELEASE = 1<<5, 1368 IFF_DONT_BRIDGE = 1<<6, 1369 IFF_DISABLE_NETPOLL = 1<<7, 1370 IFF_MACVLAN_PORT = 1<<8, 1371 IFF_BRIDGE_PORT = 1<<9, 1372 IFF_OVS_DATAPATH = 1<<10, 1373 IFF_TX_SKB_SHARING = 1<<11, 1374 IFF_UNICAST_FLT = 1<<12, 1375 IFF_TEAM_PORT = 1<<13, 1376 IFF_SUPP_NOFCS = 1<<14, 1377 IFF_LIVE_ADDR_CHANGE = 1<<15, 1378 IFF_MACVLAN = 1<<16, 1379 IFF_XMIT_DST_RELEASE_PERM = 1<<17, 1380 IFF_IPVLAN_MASTER = 1<<18, 1381 IFF_IPVLAN_SLAVE = 1<<19, 1382 IFF_L3MDEV_MASTER = 1<<20, 1383 IFF_NO_QUEUE = 1<<21, 1384 IFF_OPENVSWITCH = 1<<22, 1385 IFF_L3MDEV_SLAVE = 1<<23, 1386 IFF_TEAM = 1<<24, 1387 IFF_RXFH_CONFIGURED = 1<<25, 1388 IFF_PHONY_HEADROOM = 1<<26, 1389 IFF_MACSEC = 1<<27, 1390 }; 1391 1392 #define IFF_802_1Q_VLAN IFF_802_1Q_VLAN 1393 #define IFF_EBRIDGE IFF_EBRIDGE 1394 #define IFF_BONDING IFF_BONDING 1395 #define IFF_ISATAP IFF_ISATAP 1396 #define IFF_WAN_HDLC IFF_WAN_HDLC 1397 #define IFF_XMIT_DST_RELEASE IFF_XMIT_DST_RELEASE 1398 #define IFF_DONT_BRIDGE IFF_DONT_BRIDGE 1399 #define IFF_DISABLE_NETPOLL IFF_DISABLE_NETPOLL 1400 #define IFF_MACVLAN_PORT IFF_MACVLAN_PORT 1401 #define IFF_BRIDGE_PORT IFF_BRIDGE_PORT 1402 #define IFF_OVS_DATAPATH IFF_OVS_DATAPATH 1403 #define IFF_TX_SKB_SHARING IFF_TX_SKB_SHARING 1404 #define IFF_UNICAST_FLT IFF_UNICAST_FLT 1405 #define IFF_TEAM_PORT IFF_TEAM_PORT 1406 #define IFF_SUPP_NOFCS IFF_SUPP_NOFCS 1407 #define IFF_LIVE_ADDR_CHANGE IFF_LIVE_ADDR_CHANGE 1408 #define IFF_MACVLAN IFF_MACVLAN 1409 #define IFF_XMIT_DST_RELEASE_PERM IFF_XMIT_DST_RELEASE_PERM 1410 #define IFF_IPVLAN_MASTER IFF_IPVLAN_MASTER 1411 #define IFF_IPVLAN_SLAVE IFF_IPVLAN_SLAVE 1412 #define IFF_L3MDEV_MASTER IFF_L3MDEV_MASTER 1413 #define IFF_NO_QUEUE IFF_NO_QUEUE 1414 #define IFF_OPENVSWITCH IFF_OPENVSWITCH 1415 #define IFF_L3MDEV_SLAVE IFF_L3MDEV_SLAVE 1416 #define IFF_TEAM IFF_TEAM 1417 #define IFF_RXFH_CONFIGURED IFF_RXFH_CONFIGURED 1418 #define IFF_MACSEC IFF_MACSEC 1419 1420 /** 1421 * struct net_device - The DEVICE structure. 1422 * 1423 * Actually, this whole structure is a big mistake. It mixes I/O 1424 * data with strictly "high-level" data, and it has to know about 1425 * almost every data structure used in the INET module. 1426 * 1427 * @name: This is the first field of the "visible" part of this structure 1428 * (i.e. as seen by users in the "Space.c" file). It is the name 1429 * of the interface. 1430 * 1431 * @name_hlist: Device name hash chain, please keep it close to name[] 1432 * @ifalias: SNMP alias 1433 * @mem_end: Shared memory end 1434 * @mem_start: Shared memory start 1435 * @base_addr: Device I/O address 1436 * @irq: Device IRQ number 1437 * 1438 * @carrier_changes: Stats to monitor carrier on<->off transitions 1439 * 1440 * @state: Generic network queuing layer state, see netdev_state_t 1441 * @dev_list: The global list of network devices 1442 * @napi_list: List entry used for polling NAPI devices 1443 * @unreg_list: List entry when we are unregistering the 1444 * device; see the function unregister_netdev 1445 * @close_list: List entry used when we are closing the device 1446 * @ptype_all: Device-specific packet handlers for all protocols 1447 * @ptype_specific: Device-specific, protocol-specific packet handlers 1448 * 1449 * @adj_list: Directly linked devices, like slaves for bonding 1450 * @features: Currently active device features 1451 * @hw_features: User-changeable features 1452 * 1453 * @wanted_features: User-requested features 1454 * @vlan_features: Mask of features inheritable by VLAN devices 1455 * 1456 * @hw_enc_features: Mask of features inherited by encapsulating devices 1457 * This field indicates what encapsulation 1458 * offloads the hardware is capable of doing, 1459 * and drivers will need to set them appropriately. 1460 * 1461 * @mpls_features: Mask of features inheritable by MPLS 1462 * 1463 * @ifindex: interface index 1464 * @group: The group the device belongs to 1465 * 1466 * @stats: Statistics struct, which was left as a legacy, use 1467 * rtnl_link_stats64 instead 1468 * 1469 * @rx_dropped: Dropped packets by core network, 1470 * do not use this in drivers 1471 * @tx_dropped: Dropped packets by core network, 1472 * do not use this in drivers 1473 * @rx_nohandler: nohandler dropped packets by core network on 1474 * inactive devices, do not use this in drivers 1475 * 1476 * @wireless_handlers: List of functions to handle Wireless Extensions, 1477 * instead of ioctl, 1478 * see <net/iw_handler.h> for details. 1479 * @wireless_data: Instance data managed by the core of wireless extensions 1480 * 1481 * @netdev_ops: Includes several pointers to callbacks, 1482 * if one wants to override the ndo_*() functions 1483 * @ethtool_ops: Management operations 1484 * @ndisc_ops: Includes callbacks for different IPv6 neighbour 1485 * discovery handling. Necessary for e.g. 6LoWPAN. 1486 * @header_ops: Includes callbacks for creating,parsing,caching,etc 1487 * of Layer 2 headers. 1488 * 1489 * @flags: Interface flags (a la BSD) 1490 * @priv_flags: Like 'flags' but invisible to userspace, 1491 * see if.h for the definitions 1492 * @gflags: Global flags ( kept as legacy ) 1493 * @padded: How much padding added by alloc_netdev() 1494 * @operstate: RFC2863 operstate 1495 * @link_mode: Mapping policy to operstate 1496 * @if_port: Selectable AUI, TP, ... 1497 * @dma: DMA channel 1498 * @mtu: Interface MTU value 1499 * @min_mtu: Interface Minimum MTU value 1500 * @max_mtu: Interface Maximum MTU value 1501 * @type: Interface hardware type 1502 * @hard_header_len: Maximum hardware header length. 1503 * @min_header_len: Minimum hardware header length 1504 * 1505 * @needed_headroom: Extra headroom the hardware may need, but not in all 1506 * cases can this be guaranteed 1507 * @needed_tailroom: Extra tailroom the hardware may need, but not in all 1508 * cases can this be guaranteed. Some cases also use 1509 * LL_MAX_HEADER instead to allocate the skb 1510 * 1511 * interface address info: 1512 * 1513 * @perm_addr: Permanent hw address 1514 * @addr_assign_type: Hw address assignment type 1515 * @addr_len: Hardware address length 1516 * @neigh_priv_len: Used in neigh_alloc() 1517 * @dev_id: Used to differentiate devices that share 1518 * the same link layer address 1519 * @dev_port: Used to differentiate devices that share 1520 * the same function 1521 * @addr_list_lock: XXX: need comments on this one 1522 * @uc_promisc: Counter that indicates promiscuous mode 1523 * has been enabled due to the need to listen to 1524 * additional unicast addresses in a device that 1525 * does not implement ndo_set_rx_mode() 1526 * @uc: unicast mac addresses 1527 * @mc: multicast mac addresses 1528 * @dev_addrs: list of device hw addresses 1529 * @queues_kset: Group of all Kobjects in the Tx and RX queues 1530 * @promiscuity: Number of times the NIC is told to work in 1531 * promiscuous mode; if it becomes 0 the NIC will 1532 * exit promiscuous mode 1533 * @allmulti: Counter, enables or disables allmulticast mode 1534 * 1535 * @vlan_info: VLAN info 1536 * @dsa_ptr: dsa specific data 1537 * @tipc_ptr: TIPC specific data 1538 * @atalk_ptr: AppleTalk link 1539 * @ip_ptr: IPv4 specific data 1540 * @dn_ptr: DECnet specific data 1541 * @ip6_ptr: IPv6 specific data 1542 * @ax25_ptr: AX.25 specific data 1543 * @ieee80211_ptr: IEEE 802.11 specific data, assign before registering 1544 * 1545 * @dev_addr: Hw address (before bcast, 1546 * because most packets are unicast) 1547 * 1548 * @_rx: Array of RX queues 1549 * @num_rx_queues: Number of RX queues 1550 * allocated at register_netdev() time 1551 * @real_num_rx_queues: Number of RX queues currently active in device 1552 * 1553 * @rx_handler: handler for received packets 1554 * @rx_handler_data: XXX: need comments on this one 1555 * @ingress_queue: XXX: need comments on this one 1556 * @broadcast: hw bcast address 1557 * 1558 * @rx_cpu_rmap: CPU reverse-mapping for RX completion interrupts, 1559 * indexed by RX queue number. Assigned by driver. 1560 * This must only be set if the ndo_rx_flow_steer 1561 * operation is defined 1562 * @index_hlist: Device index hash chain 1563 * 1564 * @_tx: Array of TX queues 1565 * @num_tx_queues: Number of TX queues allocated at alloc_netdev_mq() time 1566 * @real_num_tx_queues: Number of TX queues currently active in device 1567 * @qdisc: Root qdisc from userspace point of view 1568 * @tx_queue_len: Max frames per queue allowed 1569 * @tx_global_lock: XXX: need comments on this one 1570 * 1571 * @xps_maps: XXX: need comments on this one 1572 * 1573 * @watchdog_timeo: Represents the timeout that is used by 1574 * the watchdog (see dev_watchdog()) 1575 * @watchdog_timer: List of timers 1576 * 1577 * @pcpu_refcnt: Number of references to this device 1578 * @todo_list: Delayed register/unregister 1579 * @link_watch_list: XXX: need comments on this one 1580 * 1581 * @reg_state: Register/unregister state machine 1582 * @dismantle: Device is going to be freed 1583 * @rtnl_link_state: This enum represents the phases of creating 1584 * a new link 1585 * 1586 * @needs_free_netdev: Should unregister perform free_netdev? 1587 * @priv_destructor: Called from unregister 1588 * @npinfo: XXX: need comments on this one 1589 * @nd_net: Network namespace this network device is inside 1590 * 1591 * @ml_priv: Mid-layer private 1592 * @lstats: Loopback statistics 1593 * @tstats: Tunnel statistics 1594 * @dstats: Dummy statistics 1595 * @vstats: Virtual ethernet statistics 1596 * 1597 * @garp_port: GARP 1598 * @mrp_port: MRP 1599 * 1600 * @dev: Class/net/name entry 1601 * @sysfs_groups: Space for optional device, statistics and wireless 1602 * sysfs groups 1603 * 1604 * @sysfs_rx_queue_group: Space for optional per-rx queue attributes 1605 * @rtnl_link_ops: Rtnl_link_ops 1606 * 1607 * @gso_max_size: Maximum size of generic segmentation offload 1608 * @gso_max_segs: Maximum number of segments that can be passed to the 1609 * NIC for GSO 1610 * 1611 * @dcbnl_ops: Data Center Bridging netlink ops 1612 * @num_tc: Number of traffic classes in the net device 1613 * @tc_to_txq: XXX: need comments on this one 1614 * @prio_tc_map: XXX: need comments on this one 1615 * 1616 * @fcoe_ddp_xid: Max exchange id for FCoE LRO by ddp 1617 * 1618 * @priomap: XXX: need comments on this one 1619 * @phydev: Physical device may attach itself 1620 * for hardware timestamping 1621 * 1622 * @qdisc_tx_busylock: lockdep class annotating Qdisc->busylock spinlock 1623 * @qdisc_running_key: lockdep class annotating Qdisc->running seqcount 1624 * 1625 * @proto_down: protocol port state information can be sent to the 1626 * switch driver and used to set the phys state of the 1627 * switch port. 1628 * 1629 * FIXME: cleanup struct net_device such that network protocol info 1630 * moves out. 1631 */ 1632 1633 struct net_device { 1634 char name[IFNAMSIZ]; 1635 struct hlist_node name_hlist; 1636 char *ifalias; 1637 /* 1638 * I/O specific fields 1639 * FIXME: Merge these and struct ifmap into one 1640 */ 1641 unsigned long mem_end; 1642 unsigned long mem_start; 1643 unsigned long base_addr; 1644 int irq; 1645 1646 atomic_t carrier_changes; 1647 1648 /* 1649 * Some hardware also needs these fields (state,dev_list, 1650 * napi_list,unreg_list,close_list) but they are not 1651 * part of the usual set specified in Space.c. 1652 */ 1653 1654 unsigned long state; 1655 1656 struct list_head dev_list; 1657 struct list_head napi_list; 1658 struct list_head unreg_list; 1659 struct list_head close_list; 1660 struct list_head ptype_all; 1661 struct list_head ptype_specific; 1662 1663 struct { 1664 struct list_head upper; 1665 struct list_head lower; 1666 } adj_list; 1667 1668 netdev_features_t features; 1669 netdev_features_t hw_features; 1670 netdev_features_t wanted_features; 1671 netdev_features_t vlan_features; 1672 netdev_features_t hw_enc_features; 1673 netdev_features_t mpls_features; 1674 netdev_features_t gso_partial_features; 1675 1676 int ifindex; 1677 int group; 1678 1679 struct net_device_stats stats; 1680 1681 atomic_long_t rx_dropped; 1682 atomic_long_t tx_dropped; 1683 atomic_long_t rx_nohandler; 1684 1685 #ifdef CONFIG_WIRELESS_EXT 1686 const struct iw_handler_def *wireless_handlers; 1687 struct iw_public_data *wireless_data; 1688 #endif 1689 const struct net_device_ops *netdev_ops; 1690 const struct ethtool_ops *ethtool_ops; 1691 #ifdef CONFIG_NET_SWITCHDEV 1692 const struct switchdev_ops *switchdev_ops; 1693 #endif 1694 #ifdef CONFIG_NET_L3_MASTER_DEV 1695 const struct l3mdev_ops *l3mdev_ops; 1696 #endif 1697 #if IS_ENABLED(CONFIG_IPV6) 1698 const struct ndisc_ops *ndisc_ops; 1699 #endif 1700 1701 #ifdef CONFIG_XFRM 1702 const struct xfrmdev_ops *xfrmdev_ops; 1703 #endif 1704 1705 const struct header_ops *header_ops; 1706 1707 unsigned int flags; 1708 unsigned int priv_flags; 1709 1710 unsigned short gflags; 1711 unsigned short padded; 1712 1713 unsigned char operstate; 1714 unsigned char link_mode; 1715 1716 unsigned char if_port; 1717 unsigned char dma; 1718 1719 unsigned int mtu; 1720 unsigned int min_mtu; 1721 unsigned int max_mtu; 1722 unsigned short type; 1723 unsigned short hard_header_len; 1724 unsigned char min_header_len; 1725 1726 unsigned short needed_headroom; 1727 unsigned short needed_tailroom; 1728 1729 /* Interface address info. */ 1730 unsigned char perm_addr[MAX_ADDR_LEN]; 1731 unsigned char addr_assign_type; 1732 unsigned char addr_len; 1733 unsigned short neigh_priv_len; 1734 unsigned short dev_id; 1735 unsigned short dev_port; 1736 spinlock_t addr_list_lock; 1737 unsigned char name_assign_type; 1738 bool uc_promisc; 1739 struct netdev_hw_addr_list uc; 1740 struct netdev_hw_addr_list mc; 1741 struct netdev_hw_addr_list dev_addrs; 1742 1743 #ifdef CONFIG_SYSFS 1744 struct kset *queues_kset; 1745 #endif 1746 unsigned int promiscuity; 1747 unsigned int allmulti; 1748 1749 1750 /* Protocol-specific pointers */ 1751 1752 #if IS_ENABLED(CONFIG_VLAN_8021Q) 1753 struct vlan_info __rcu *vlan_info; 1754 #endif 1755 #if IS_ENABLED(CONFIG_NET_DSA) 1756 struct dsa_switch_tree *dsa_ptr; 1757 #endif 1758 #if IS_ENABLED(CONFIG_TIPC) 1759 struct tipc_bearer __rcu *tipc_ptr; 1760 #endif 1761 void *atalk_ptr; 1762 struct in_device __rcu *ip_ptr; 1763 struct dn_dev __rcu *dn_ptr; 1764 struct inet6_dev __rcu *ip6_ptr; 1765 void *ax25_ptr; 1766 struct wireless_dev *ieee80211_ptr; 1767 struct wpan_dev *ieee802154_ptr; 1768 #if IS_ENABLED(CONFIG_MPLS_ROUTING) 1769 struct mpls_dev __rcu *mpls_ptr; 1770 #endif 1771 1772 /* 1773 * Cache lines mostly used on receive path (including eth_type_trans()) 1774 */ 1775 /* Interface address info used in eth_type_trans() */ 1776 unsigned char *dev_addr; 1777 1778 #ifdef CONFIG_SYSFS 1779 struct netdev_rx_queue *_rx; 1780 1781 unsigned int num_rx_queues; 1782 unsigned int real_num_rx_queues; 1783 #endif 1784 1785 struct bpf_prog __rcu *xdp_prog; 1786 unsigned long gro_flush_timeout; 1787 rx_handler_func_t __rcu *rx_handler; 1788 void __rcu *rx_handler_data; 1789 1790 #ifdef CONFIG_NET_CLS_ACT 1791 struct tcf_proto __rcu *ingress_cl_list; 1792 #endif 1793 struct netdev_queue __rcu *ingress_queue; 1794 #ifdef CONFIG_NETFILTER_INGRESS 1795 struct nf_hook_entry __rcu *nf_hooks_ingress; 1796 #endif 1797 1798 unsigned char broadcast[MAX_ADDR_LEN]; 1799 #ifdef CONFIG_RFS_ACCEL 1800 struct cpu_rmap *rx_cpu_rmap; 1801 #endif 1802 struct hlist_node index_hlist; 1803 1804 /* 1805 * Cache lines mostly used on transmit path 1806 */ 1807 struct netdev_queue *_tx ____cacheline_aligned_in_smp; 1808 unsigned int num_tx_queues; 1809 unsigned int real_num_tx_queues; 1810 struct Qdisc *qdisc; 1811 #ifdef CONFIG_NET_SCHED 1812 DECLARE_HASHTABLE (qdisc_hash, 4); 1813 #endif 1814 unsigned int tx_queue_len; 1815 spinlock_t tx_global_lock; 1816 int watchdog_timeo; 1817 1818 #ifdef CONFIG_XPS 1819 struct xps_dev_maps __rcu *xps_maps; 1820 #endif 1821 #ifdef CONFIG_NET_CLS_ACT 1822 struct tcf_proto __rcu *egress_cl_list; 1823 #endif 1824 1825 /* These may be needed for future network-power-down code. */ 1826 struct timer_list watchdog_timer; 1827 1828 int __percpu *pcpu_refcnt; 1829 struct list_head todo_list; 1830 1831 struct list_head link_watch_list; 1832 1833 enum { NETREG_UNINITIALIZED=0, 1834 NETREG_REGISTERED, /* completed register_netdevice */ 1835 NETREG_UNREGISTERING, /* called unregister_netdevice */ 1836 NETREG_UNREGISTERED, /* completed unregister todo */ 1837 NETREG_RELEASED, /* called free_netdev */ 1838 NETREG_DUMMY, /* dummy device for NAPI poll */ 1839 } reg_state:8; 1840 1841 bool dismantle; 1842 1843 enum { 1844 RTNL_LINK_INITIALIZED, 1845 RTNL_LINK_INITIALIZING, 1846 } rtnl_link_state:16; 1847 1848 bool needs_free_netdev; 1849 void (*priv_destructor)(struct net_device *dev); 1850 1851 #ifdef CONFIG_NETPOLL 1852 struct netpoll_info __rcu *npinfo; 1853 #endif 1854 1855 possible_net_t nd_net; 1856 1857 /* mid-layer private */ 1858 union { 1859 void *ml_priv; 1860 struct pcpu_lstats __percpu *lstats; 1861 struct pcpu_sw_netstats __percpu *tstats; 1862 struct pcpu_dstats __percpu *dstats; 1863 struct pcpu_vstats __percpu *vstats; 1864 }; 1865 1866 #if IS_ENABLED(CONFIG_GARP) 1867 struct garp_port __rcu *garp_port; 1868 #endif 1869 #if IS_ENABLED(CONFIG_MRP) 1870 struct mrp_port __rcu *mrp_port; 1871 #endif 1872 1873 struct device dev; 1874 const struct attribute_group *sysfs_groups[4]; 1875 const struct attribute_group *sysfs_rx_queue_group; 1876 1877 const struct rtnl_link_ops *rtnl_link_ops; 1878 1879 /* for setting kernel sock attribute on TCP connection setup */ 1880 #define GSO_MAX_SIZE 65536 1881 unsigned int gso_max_size; 1882 #define GSO_MAX_SEGS 65535 1883 u16 gso_max_segs; 1884 1885 #ifdef CONFIG_DCB 1886 const struct dcbnl_rtnl_ops *dcbnl_ops; 1887 #endif 1888 u8 num_tc; 1889 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE]; 1890 u8 prio_tc_map[TC_BITMASK + 1]; 1891 1892 #if IS_ENABLED(CONFIG_FCOE) 1893 unsigned int fcoe_ddp_xid; 1894 #endif 1895 #if IS_ENABLED(CONFIG_CGROUP_NET_PRIO) 1896 struct netprio_map __rcu *priomap; 1897 #endif 1898 struct phy_device *phydev; 1899 struct lock_class_key *qdisc_tx_busylock; 1900 struct lock_class_key *qdisc_running_key; 1901 bool proto_down; 1902 }; 1903 #define to_net_dev(d) container_of(d, struct net_device, dev) 1904 1905 static inline bool netif_elide_gro(const struct net_device *dev) 1906 { 1907 if (!(dev->features & NETIF_F_GRO) || dev->xdp_prog) 1908 return true; 1909 return false; 1910 } 1911 1912 #define NETDEV_ALIGN 32 1913 1914 static inline 1915 int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio) 1916 { 1917 return dev->prio_tc_map[prio & TC_BITMASK]; 1918 } 1919 1920 static inline 1921 int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc) 1922 { 1923 if (tc >= dev->num_tc) 1924 return -EINVAL; 1925 1926 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK; 1927 return 0; 1928 } 1929 1930 int netdev_txq_to_tc(struct net_device *dev, unsigned int txq); 1931 void netdev_reset_tc(struct net_device *dev); 1932 int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset); 1933 int netdev_set_num_tc(struct net_device *dev, u8 num_tc); 1934 1935 static inline 1936 int netdev_get_num_tc(struct net_device *dev) 1937 { 1938 return dev->num_tc; 1939 } 1940 1941 static inline 1942 struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev, 1943 unsigned int index) 1944 { 1945 return &dev->_tx[index]; 1946 } 1947 1948 static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev, 1949 const struct sk_buff *skb) 1950 { 1951 return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb)); 1952 } 1953 1954 static inline void netdev_for_each_tx_queue(struct net_device *dev, 1955 void (*f)(struct net_device *, 1956 struct netdev_queue *, 1957 void *), 1958 void *arg) 1959 { 1960 unsigned int i; 1961 1962 for (i = 0; i < dev->num_tx_queues; i++) 1963 f(dev, &dev->_tx[i], arg); 1964 } 1965 1966 #define netdev_lockdep_set_classes(dev) \ 1967 { \ 1968 static struct lock_class_key qdisc_tx_busylock_key; \ 1969 static struct lock_class_key qdisc_running_key; \ 1970 static struct lock_class_key qdisc_xmit_lock_key; \ 1971 static struct lock_class_key dev_addr_list_lock_key; \ 1972 unsigned int i; \ 1973 \ 1974 (dev)->qdisc_tx_busylock = &qdisc_tx_busylock_key; \ 1975 (dev)->qdisc_running_key = &qdisc_running_key; \ 1976 lockdep_set_class(&(dev)->addr_list_lock, \ 1977 &dev_addr_list_lock_key); \ 1978 for (i = 0; i < (dev)->num_tx_queues; i++) \ 1979 lockdep_set_class(&(dev)->_tx[i]._xmit_lock, \ 1980 &qdisc_xmit_lock_key); \ 1981 } 1982 1983 struct netdev_queue *netdev_pick_tx(struct net_device *dev, 1984 struct sk_buff *skb, 1985 void *accel_priv); 1986 1987 /* returns the headroom that the master device needs to take in account 1988 * when forwarding to this dev 1989 */ 1990 static inline unsigned netdev_get_fwd_headroom(struct net_device *dev) 1991 { 1992 return dev->priv_flags & IFF_PHONY_HEADROOM ? 0 : dev->needed_headroom; 1993 } 1994 1995 static inline void netdev_set_rx_headroom(struct net_device *dev, int new_hr) 1996 { 1997 if (dev->netdev_ops->ndo_set_rx_headroom) 1998 dev->netdev_ops->ndo_set_rx_headroom(dev, new_hr); 1999 } 2000 2001 /* set the device rx headroom to the dev's default */ 2002 static inline void netdev_reset_rx_headroom(struct net_device *dev) 2003 { 2004 netdev_set_rx_headroom(dev, -1); 2005 } 2006 2007 /* 2008 * Net namespace inlines 2009 */ 2010 static inline 2011 struct net *dev_net(const struct net_device *dev) 2012 { 2013 return read_pnet(&dev->nd_net); 2014 } 2015 2016 static inline 2017 void dev_net_set(struct net_device *dev, struct net *net) 2018 { 2019 write_pnet(&dev->nd_net, net); 2020 } 2021 2022 /** 2023 * netdev_priv - access network device private data 2024 * @dev: network device 2025 * 2026 * Get network device private data 2027 */ 2028 static inline void *netdev_priv(const struct net_device *dev) 2029 { 2030 return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN); 2031 } 2032 2033 /* Set the sysfs physical device reference for the network logical device 2034 * if set prior to registration will cause a symlink during initialization. 2035 */ 2036 #define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev)) 2037 2038 /* Set the sysfs device type for the network logical device to allow 2039 * fine-grained identification of different network device types. For 2040 * example Ethernet, Wireless LAN, Bluetooth, WiMAX etc. 2041 */ 2042 #define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype)) 2043 2044 /* Default NAPI poll() weight 2045 * Device drivers are strongly advised to not use bigger value 2046 */ 2047 #define NAPI_POLL_WEIGHT 64 2048 2049 /** 2050 * netif_napi_add - initialize a NAPI context 2051 * @dev: network device 2052 * @napi: NAPI context 2053 * @poll: polling function 2054 * @weight: default weight 2055 * 2056 * netif_napi_add() must be used to initialize a NAPI context prior to calling 2057 * *any* of the other NAPI-related functions. 2058 */ 2059 void netif_napi_add(struct net_device *dev, struct napi_struct *napi, 2060 int (*poll)(struct napi_struct *, int), int weight); 2061 2062 /** 2063 * netif_tx_napi_add - initialize a NAPI context 2064 * @dev: network device 2065 * @napi: NAPI context 2066 * @poll: polling function 2067 * @weight: default weight 2068 * 2069 * This variant of netif_napi_add() should be used from drivers using NAPI 2070 * to exclusively poll a TX queue. 2071 * This will avoid we add it into napi_hash[], thus polluting this hash table. 2072 */ 2073 static inline void netif_tx_napi_add(struct net_device *dev, 2074 struct napi_struct *napi, 2075 int (*poll)(struct napi_struct *, int), 2076 int weight) 2077 { 2078 set_bit(NAPI_STATE_NO_BUSY_POLL, &napi->state); 2079 netif_napi_add(dev, napi, poll, weight); 2080 } 2081 2082 /** 2083 * netif_napi_del - remove a NAPI context 2084 * @napi: NAPI context 2085 * 2086 * netif_napi_del() removes a NAPI context from the network device NAPI list 2087 */ 2088 void netif_napi_del(struct napi_struct *napi); 2089 2090 struct napi_gro_cb { 2091 /* Virtual address of skb_shinfo(skb)->frags[0].page + offset. */ 2092 void *frag0; 2093 2094 /* Length of frag0. */ 2095 unsigned int frag0_len; 2096 2097 /* This indicates where we are processing relative to skb->data. */ 2098 int data_offset; 2099 2100 /* This is non-zero if the packet cannot be merged with the new skb. */ 2101 u16 flush; 2102 2103 /* Save the IP ID here and check when we get to the transport layer */ 2104 u16 flush_id; 2105 2106 /* Number of segments aggregated. */ 2107 u16 count; 2108 2109 /* Start offset for remote checksum offload */ 2110 u16 gro_remcsum_start; 2111 2112 /* jiffies when first packet was created/queued */ 2113 unsigned long age; 2114 2115 /* Used in ipv6_gro_receive() and foo-over-udp */ 2116 u16 proto; 2117 2118 /* This is non-zero if the packet may be of the same flow. */ 2119 u8 same_flow:1; 2120 2121 /* Used in tunnel GRO receive */ 2122 u8 encap_mark:1; 2123 2124 /* GRO checksum is valid */ 2125 u8 csum_valid:1; 2126 2127 /* Number of checksums via CHECKSUM_UNNECESSARY */ 2128 u8 csum_cnt:3; 2129 2130 /* Free the skb? */ 2131 u8 free:2; 2132 #define NAPI_GRO_FREE 1 2133 #define NAPI_GRO_FREE_STOLEN_HEAD 2 2134 2135 /* Used in foo-over-udp, set in udp[46]_gro_receive */ 2136 u8 is_ipv6:1; 2137 2138 /* Used in GRE, set in fou/gue_gro_receive */ 2139 u8 is_fou:1; 2140 2141 /* Used to determine if flush_id can be ignored */ 2142 u8 is_atomic:1; 2143 2144 /* Number of gro_receive callbacks this packet already went through */ 2145 u8 recursion_counter:4; 2146 2147 /* 1 bit hole */ 2148 2149 /* used to support CHECKSUM_COMPLETE for tunneling protocols */ 2150 __wsum csum; 2151 2152 /* used in skb_gro_receive() slow path */ 2153 struct sk_buff *last; 2154 }; 2155 2156 #define NAPI_GRO_CB(skb) ((struct napi_gro_cb *)(skb)->cb) 2157 2158 #define GRO_RECURSION_LIMIT 15 2159 static inline int gro_recursion_inc_test(struct sk_buff *skb) 2160 { 2161 return ++NAPI_GRO_CB(skb)->recursion_counter == GRO_RECURSION_LIMIT; 2162 } 2163 2164 typedef struct sk_buff **(*gro_receive_t)(struct sk_buff **, struct sk_buff *); 2165 static inline struct sk_buff **call_gro_receive(gro_receive_t cb, 2166 struct sk_buff **head, 2167 struct sk_buff *skb) 2168 { 2169 if (unlikely(gro_recursion_inc_test(skb))) { 2170 NAPI_GRO_CB(skb)->flush |= 1; 2171 return NULL; 2172 } 2173 2174 return cb(head, skb); 2175 } 2176 2177 typedef struct sk_buff **(*gro_receive_sk_t)(struct sock *, struct sk_buff **, 2178 struct sk_buff *); 2179 static inline struct sk_buff **call_gro_receive_sk(gro_receive_sk_t cb, 2180 struct sock *sk, 2181 struct sk_buff **head, 2182 struct sk_buff *skb) 2183 { 2184 if (unlikely(gro_recursion_inc_test(skb))) { 2185 NAPI_GRO_CB(skb)->flush |= 1; 2186 return NULL; 2187 } 2188 2189 return cb(sk, head, skb); 2190 } 2191 2192 struct packet_type { 2193 __be16 type; /* This is really htons(ether_type). */ 2194 struct net_device *dev; /* NULL is wildcarded here */ 2195 int (*func) (struct sk_buff *, 2196 struct net_device *, 2197 struct packet_type *, 2198 struct net_device *); 2199 bool (*id_match)(struct packet_type *ptype, 2200 struct sock *sk); 2201 void *af_packet_priv; 2202 struct list_head list; 2203 }; 2204 2205 struct offload_callbacks { 2206 struct sk_buff *(*gso_segment)(struct sk_buff *skb, 2207 netdev_features_t features); 2208 struct sk_buff **(*gro_receive)(struct sk_buff **head, 2209 struct sk_buff *skb); 2210 int (*gro_complete)(struct sk_buff *skb, int nhoff); 2211 }; 2212 2213 struct packet_offload { 2214 __be16 type; /* This is really htons(ether_type). */ 2215 u16 priority; 2216 struct offload_callbacks callbacks; 2217 struct list_head list; 2218 }; 2219 2220 /* often modified stats are per-CPU, other are shared (netdev->stats) */ 2221 struct pcpu_sw_netstats { 2222 u64 rx_packets; 2223 u64 rx_bytes; 2224 u64 tx_packets; 2225 u64 tx_bytes; 2226 struct u64_stats_sync syncp; 2227 }; 2228 2229 #define __netdev_alloc_pcpu_stats(type, gfp) \ 2230 ({ \ 2231 typeof(type) __percpu *pcpu_stats = alloc_percpu_gfp(type, gfp);\ 2232 if (pcpu_stats) { \ 2233 int __cpu; \ 2234 for_each_possible_cpu(__cpu) { \ 2235 typeof(type) *stat; \ 2236 stat = per_cpu_ptr(pcpu_stats, __cpu); \ 2237 u64_stats_init(&stat->syncp); \ 2238 } \ 2239 } \ 2240 pcpu_stats; \ 2241 }) 2242 2243 #define netdev_alloc_pcpu_stats(type) \ 2244 __netdev_alloc_pcpu_stats(type, GFP_KERNEL) 2245 2246 enum netdev_lag_tx_type { 2247 NETDEV_LAG_TX_TYPE_UNKNOWN, 2248 NETDEV_LAG_TX_TYPE_RANDOM, 2249 NETDEV_LAG_TX_TYPE_BROADCAST, 2250 NETDEV_LAG_TX_TYPE_ROUNDROBIN, 2251 NETDEV_LAG_TX_TYPE_ACTIVEBACKUP, 2252 NETDEV_LAG_TX_TYPE_HASH, 2253 }; 2254 2255 struct netdev_lag_upper_info { 2256 enum netdev_lag_tx_type tx_type; 2257 }; 2258 2259 struct netdev_lag_lower_state_info { 2260 u8 link_up : 1, 2261 tx_enabled : 1; 2262 }; 2263 2264 #include <linux/notifier.h> 2265 2266 /* netdevice notifier chain. Please remember to update the rtnetlink 2267 * notification exclusion list in rtnetlink_event() when adding new 2268 * types. 2269 */ 2270 #define NETDEV_UP 0x0001 /* For now you can't veto a device up/down */ 2271 #define NETDEV_DOWN 0x0002 2272 #define NETDEV_REBOOT 0x0003 /* Tell a protocol stack a network interface 2273 detected a hardware crash and restarted 2274 - we can use this eg to kick tcp sessions 2275 once done */ 2276 #define NETDEV_CHANGE 0x0004 /* Notify device state change */ 2277 #define NETDEV_REGISTER 0x0005 2278 #define NETDEV_UNREGISTER 0x0006 2279 #define NETDEV_CHANGEMTU 0x0007 /* notify after mtu change happened */ 2280 #define NETDEV_CHANGEADDR 0x0008 2281 #define NETDEV_GOING_DOWN 0x0009 2282 #define NETDEV_CHANGENAME 0x000A 2283 #define NETDEV_FEAT_CHANGE 0x000B 2284 #define NETDEV_BONDING_FAILOVER 0x000C 2285 #define NETDEV_PRE_UP 0x000D 2286 #define NETDEV_PRE_TYPE_CHANGE 0x000E 2287 #define NETDEV_POST_TYPE_CHANGE 0x000F 2288 #define NETDEV_POST_INIT 0x0010 2289 #define NETDEV_UNREGISTER_FINAL 0x0011 2290 #define NETDEV_RELEASE 0x0012 2291 #define NETDEV_NOTIFY_PEERS 0x0013 2292 #define NETDEV_JOIN 0x0014 2293 #define NETDEV_CHANGEUPPER 0x0015 2294 #define NETDEV_RESEND_IGMP 0x0016 2295 #define NETDEV_PRECHANGEMTU 0x0017 /* notify before mtu change happened */ 2296 #define NETDEV_CHANGEINFODATA 0x0018 2297 #define NETDEV_BONDING_INFO 0x0019 2298 #define NETDEV_PRECHANGEUPPER 0x001A 2299 #define NETDEV_CHANGELOWERSTATE 0x001B 2300 #define NETDEV_UDP_TUNNEL_PUSH_INFO 0x001C 2301 #define NETDEV_UDP_TUNNEL_DROP_INFO 0x001D 2302 #define NETDEV_CHANGE_TX_QUEUE_LEN 0x001E 2303 2304 int register_netdevice_notifier(struct notifier_block *nb); 2305 int unregister_netdevice_notifier(struct notifier_block *nb); 2306 2307 struct netdev_notifier_info { 2308 struct net_device *dev; 2309 }; 2310 2311 struct netdev_notifier_change_info { 2312 struct netdev_notifier_info info; /* must be first */ 2313 unsigned int flags_changed; 2314 }; 2315 2316 struct netdev_notifier_changeupper_info { 2317 struct netdev_notifier_info info; /* must be first */ 2318 struct net_device *upper_dev; /* new upper dev */ 2319 bool master; /* is upper dev master */ 2320 bool linking; /* is the notification for link or unlink */ 2321 void *upper_info; /* upper dev info */ 2322 }; 2323 2324 struct netdev_notifier_changelowerstate_info { 2325 struct netdev_notifier_info info; /* must be first */ 2326 void *lower_state_info; /* is lower dev state */ 2327 }; 2328 2329 static inline void netdev_notifier_info_init(struct netdev_notifier_info *info, 2330 struct net_device *dev) 2331 { 2332 info->dev = dev; 2333 } 2334 2335 static inline struct net_device * 2336 netdev_notifier_info_to_dev(const struct netdev_notifier_info *info) 2337 { 2338 return info->dev; 2339 } 2340 2341 int call_netdevice_notifiers(unsigned long val, struct net_device *dev); 2342 2343 2344 extern rwlock_t dev_base_lock; /* Device list lock */ 2345 2346 #define for_each_netdev(net, d) \ 2347 list_for_each_entry(d, &(net)->dev_base_head, dev_list) 2348 #define for_each_netdev_reverse(net, d) \ 2349 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list) 2350 #define for_each_netdev_rcu(net, d) \ 2351 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list) 2352 #define for_each_netdev_safe(net, d, n) \ 2353 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list) 2354 #define for_each_netdev_continue(net, d) \ 2355 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list) 2356 #define for_each_netdev_continue_rcu(net, d) \ 2357 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list) 2358 #define for_each_netdev_in_bond_rcu(bond, slave) \ 2359 for_each_netdev_rcu(&init_net, slave) \ 2360 if (netdev_master_upper_dev_get_rcu(slave) == (bond)) 2361 #define net_device_entry(lh) list_entry(lh, struct net_device, dev_list) 2362 2363 static inline struct net_device *next_net_device(struct net_device *dev) 2364 { 2365 struct list_head *lh; 2366 struct net *net; 2367 2368 net = dev_net(dev); 2369 lh = dev->dev_list.next; 2370 return lh == &net->dev_base_head ? NULL : net_device_entry(lh); 2371 } 2372 2373 static inline struct net_device *next_net_device_rcu(struct net_device *dev) 2374 { 2375 struct list_head *lh; 2376 struct net *net; 2377 2378 net = dev_net(dev); 2379 lh = rcu_dereference(list_next_rcu(&dev->dev_list)); 2380 return lh == &net->dev_base_head ? NULL : net_device_entry(lh); 2381 } 2382 2383 static inline struct net_device *first_net_device(struct net *net) 2384 { 2385 return list_empty(&net->dev_base_head) ? NULL : 2386 net_device_entry(net->dev_base_head.next); 2387 } 2388 2389 static inline struct net_device *first_net_device_rcu(struct net *net) 2390 { 2391 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head)); 2392 2393 return lh == &net->dev_base_head ? NULL : net_device_entry(lh); 2394 } 2395 2396 int netdev_boot_setup_check(struct net_device *dev); 2397 unsigned long netdev_boot_base(const char *prefix, int unit); 2398 struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type, 2399 const char *hwaddr); 2400 struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type); 2401 struct net_device *__dev_getfirstbyhwtype(struct net *net, unsigned short type); 2402 void dev_add_pack(struct packet_type *pt); 2403 void dev_remove_pack(struct packet_type *pt); 2404 void __dev_remove_pack(struct packet_type *pt); 2405 void dev_add_offload(struct packet_offload *po); 2406 void dev_remove_offload(struct packet_offload *po); 2407 2408 int dev_get_iflink(const struct net_device *dev); 2409 int dev_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb); 2410 struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags, 2411 unsigned short mask); 2412 struct net_device *dev_get_by_name(struct net *net, const char *name); 2413 struct net_device *dev_get_by_name_rcu(struct net *net, const char *name); 2414 struct net_device *__dev_get_by_name(struct net *net, const char *name); 2415 int dev_alloc_name(struct net_device *dev, const char *name); 2416 int dev_open(struct net_device *dev); 2417 void dev_close(struct net_device *dev); 2418 void dev_close_many(struct list_head *head, bool unlink); 2419 void dev_disable_lro(struct net_device *dev); 2420 int dev_loopback_xmit(struct net *net, struct sock *sk, struct sk_buff *newskb); 2421 int dev_queue_xmit(struct sk_buff *skb); 2422 int dev_queue_xmit_accel(struct sk_buff *skb, void *accel_priv); 2423 int register_netdevice(struct net_device *dev); 2424 void unregister_netdevice_queue(struct net_device *dev, struct list_head *head); 2425 void unregister_netdevice_many(struct list_head *head); 2426 static inline void unregister_netdevice(struct net_device *dev) 2427 { 2428 unregister_netdevice_queue(dev, NULL); 2429 } 2430 2431 int netdev_refcnt_read(const struct net_device *dev); 2432 void free_netdev(struct net_device *dev); 2433 void netdev_freemem(struct net_device *dev); 2434 void synchronize_net(void); 2435 int init_dummy_netdev(struct net_device *dev); 2436 2437 DECLARE_PER_CPU(int, xmit_recursion); 2438 #define XMIT_RECURSION_LIMIT 10 2439 2440 static inline int dev_recursion_level(void) 2441 { 2442 return this_cpu_read(xmit_recursion); 2443 } 2444 2445 struct net_device *dev_get_by_index(struct net *net, int ifindex); 2446 struct net_device *__dev_get_by_index(struct net *net, int ifindex); 2447 struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex); 2448 struct net_device *dev_get_by_napi_id(unsigned int napi_id); 2449 int netdev_get_name(struct net *net, char *name, int ifindex); 2450 int dev_restart(struct net_device *dev); 2451 int skb_gro_receive(struct sk_buff **head, struct sk_buff *skb); 2452 2453 static inline unsigned int skb_gro_offset(const struct sk_buff *skb) 2454 { 2455 return NAPI_GRO_CB(skb)->data_offset; 2456 } 2457 2458 static inline unsigned int skb_gro_len(const struct sk_buff *skb) 2459 { 2460 return skb->len - NAPI_GRO_CB(skb)->data_offset; 2461 } 2462 2463 static inline void skb_gro_pull(struct sk_buff *skb, unsigned int len) 2464 { 2465 NAPI_GRO_CB(skb)->data_offset += len; 2466 } 2467 2468 static inline void *skb_gro_header_fast(struct sk_buff *skb, 2469 unsigned int offset) 2470 { 2471 return NAPI_GRO_CB(skb)->frag0 + offset; 2472 } 2473 2474 static inline int skb_gro_header_hard(struct sk_buff *skb, unsigned int hlen) 2475 { 2476 return NAPI_GRO_CB(skb)->frag0_len < hlen; 2477 } 2478 2479 static inline void skb_gro_frag0_invalidate(struct sk_buff *skb) 2480 { 2481 NAPI_GRO_CB(skb)->frag0 = NULL; 2482 NAPI_GRO_CB(skb)->frag0_len = 0; 2483 } 2484 2485 static inline void *skb_gro_header_slow(struct sk_buff *skb, unsigned int hlen, 2486 unsigned int offset) 2487 { 2488 if (!pskb_may_pull(skb, hlen)) 2489 return NULL; 2490 2491 skb_gro_frag0_invalidate(skb); 2492 return skb->data + offset; 2493 } 2494 2495 static inline void *skb_gro_network_header(struct sk_buff *skb) 2496 { 2497 return (NAPI_GRO_CB(skb)->frag0 ?: skb->data) + 2498 skb_network_offset(skb); 2499 } 2500 2501 static inline void skb_gro_postpull_rcsum(struct sk_buff *skb, 2502 const void *start, unsigned int len) 2503 { 2504 if (NAPI_GRO_CB(skb)->csum_valid) 2505 NAPI_GRO_CB(skb)->csum = csum_sub(NAPI_GRO_CB(skb)->csum, 2506 csum_partial(start, len, 0)); 2507 } 2508 2509 /* GRO checksum functions. These are logical equivalents of the normal 2510 * checksum functions (in skbuff.h) except that they operate on the GRO 2511 * offsets and fields in sk_buff. 2512 */ 2513 2514 __sum16 __skb_gro_checksum_complete(struct sk_buff *skb); 2515 2516 static inline bool skb_at_gro_remcsum_start(struct sk_buff *skb) 2517 { 2518 return (NAPI_GRO_CB(skb)->gro_remcsum_start == skb_gro_offset(skb)); 2519 } 2520 2521 static inline bool __skb_gro_checksum_validate_needed(struct sk_buff *skb, 2522 bool zero_okay, 2523 __sum16 check) 2524 { 2525 return ((skb->ip_summed != CHECKSUM_PARTIAL || 2526 skb_checksum_start_offset(skb) < 2527 skb_gro_offset(skb)) && 2528 !skb_at_gro_remcsum_start(skb) && 2529 NAPI_GRO_CB(skb)->csum_cnt == 0 && 2530 (!zero_okay || check)); 2531 } 2532 2533 static inline __sum16 __skb_gro_checksum_validate_complete(struct sk_buff *skb, 2534 __wsum psum) 2535 { 2536 if (NAPI_GRO_CB(skb)->csum_valid && 2537 !csum_fold(csum_add(psum, NAPI_GRO_CB(skb)->csum))) 2538 return 0; 2539 2540 NAPI_GRO_CB(skb)->csum = psum; 2541 2542 return __skb_gro_checksum_complete(skb); 2543 } 2544 2545 static inline void skb_gro_incr_csum_unnecessary(struct sk_buff *skb) 2546 { 2547 if (NAPI_GRO_CB(skb)->csum_cnt > 0) { 2548 /* Consume a checksum from CHECKSUM_UNNECESSARY */ 2549 NAPI_GRO_CB(skb)->csum_cnt--; 2550 } else { 2551 /* Update skb for CHECKSUM_UNNECESSARY and csum_level when we 2552 * verified a new top level checksum or an encapsulated one 2553 * during GRO. This saves work if we fallback to normal path. 2554 */ 2555 __skb_incr_checksum_unnecessary(skb); 2556 } 2557 } 2558 2559 #define __skb_gro_checksum_validate(skb, proto, zero_okay, check, \ 2560 compute_pseudo) \ 2561 ({ \ 2562 __sum16 __ret = 0; \ 2563 if (__skb_gro_checksum_validate_needed(skb, zero_okay, check)) \ 2564 __ret = __skb_gro_checksum_validate_complete(skb, \ 2565 compute_pseudo(skb, proto)); \ 2566 if (!__ret) \ 2567 skb_gro_incr_csum_unnecessary(skb); \ 2568 __ret; \ 2569 }) 2570 2571 #define skb_gro_checksum_validate(skb, proto, compute_pseudo) \ 2572 __skb_gro_checksum_validate(skb, proto, false, 0, compute_pseudo) 2573 2574 #define skb_gro_checksum_validate_zero_check(skb, proto, check, \ 2575 compute_pseudo) \ 2576 __skb_gro_checksum_validate(skb, proto, true, check, compute_pseudo) 2577 2578 #define skb_gro_checksum_simple_validate(skb) \ 2579 __skb_gro_checksum_validate(skb, 0, false, 0, null_compute_pseudo) 2580 2581 static inline bool __skb_gro_checksum_convert_check(struct sk_buff *skb) 2582 { 2583 return (NAPI_GRO_CB(skb)->csum_cnt == 0 && 2584 !NAPI_GRO_CB(skb)->csum_valid); 2585 } 2586 2587 static inline void __skb_gro_checksum_convert(struct sk_buff *skb, 2588 __sum16 check, __wsum pseudo) 2589 { 2590 NAPI_GRO_CB(skb)->csum = ~pseudo; 2591 NAPI_GRO_CB(skb)->csum_valid = 1; 2592 } 2593 2594 #define skb_gro_checksum_try_convert(skb, proto, check, compute_pseudo) \ 2595 do { \ 2596 if (__skb_gro_checksum_convert_check(skb)) \ 2597 __skb_gro_checksum_convert(skb, check, \ 2598 compute_pseudo(skb, proto)); \ 2599 } while (0) 2600 2601 struct gro_remcsum { 2602 int offset; 2603 __wsum delta; 2604 }; 2605 2606 static inline void skb_gro_remcsum_init(struct gro_remcsum *grc) 2607 { 2608 grc->offset = 0; 2609 grc->delta = 0; 2610 } 2611 2612 static inline void *skb_gro_remcsum_process(struct sk_buff *skb, void *ptr, 2613 unsigned int off, size_t hdrlen, 2614 int start, int offset, 2615 struct gro_remcsum *grc, 2616 bool nopartial) 2617 { 2618 __wsum delta; 2619 size_t plen = hdrlen + max_t(size_t, offset + sizeof(u16), start); 2620 2621 BUG_ON(!NAPI_GRO_CB(skb)->csum_valid); 2622 2623 if (!nopartial) { 2624 NAPI_GRO_CB(skb)->gro_remcsum_start = off + hdrlen + start; 2625 return ptr; 2626 } 2627 2628 ptr = skb_gro_header_fast(skb, off); 2629 if (skb_gro_header_hard(skb, off + plen)) { 2630 ptr = skb_gro_header_slow(skb, off + plen, off); 2631 if (!ptr) 2632 return NULL; 2633 } 2634 2635 delta = remcsum_adjust(ptr + hdrlen, NAPI_GRO_CB(skb)->csum, 2636 start, offset); 2637 2638 /* Adjust skb->csum since we changed the packet */ 2639 NAPI_GRO_CB(skb)->csum = csum_add(NAPI_GRO_CB(skb)->csum, delta); 2640 2641 grc->offset = off + hdrlen + offset; 2642 grc->delta = delta; 2643 2644 return ptr; 2645 } 2646 2647 static inline void skb_gro_remcsum_cleanup(struct sk_buff *skb, 2648 struct gro_remcsum *grc) 2649 { 2650 void *ptr; 2651 size_t plen = grc->offset + sizeof(u16); 2652 2653 if (!grc->delta) 2654 return; 2655 2656 ptr = skb_gro_header_fast(skb, grc->offset); 2657 if (skb_gro_header_hard(skb, grc->offset + sizeof(u16))) { 2658 ptr = skb_gro_header_slow(skb, plen, grc->offset); 2659 if (!ptr) 2660 return; 2661 } 2662 2663 remcsum_unadjust((__sum16 *)ptr, grc->delta); 2664 } 2665 2666 #ifdef CONFIG_XFRM_OFFLOAD 2667 static inline void skb_gro_flush_final(struct sk_buff *skb, struct sk_buff **pp, int flush) 2668 { 2669 if (PTR_ERR(pp) != -EINPROGRESS) 2670 NAPI_GRO_CB(skb)->flush |= flush; 2671 } 2672 #else 2673 static inline void skb_gro_flush_final(struct sk_buff *skb, struct sk_buff **pp, int flush) 2674 { 2675 NAPI_GRO_CB(skb)->flush |= flush; 2676 } 2677 #endif 2678 2679 static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev, 2680 unsigned short type, 2681 const void *daddr, const void *saddr, 2682 unsigned int len) 2683 { 2684 if (!dev->header_ops || !dev->header_ops->create) 2685 return 0; 2686 2687 return dev->header_ops->create(skb, dev, type, daddr, saddr, len); 2688 } 2689 2690 static inline int dev_parse_header(const struct sk_buff *skb, 2691 unsigned char *haddr) 2692 { 2693 const struct net_device *dev = skb->dev; 2694 2695 if (!dev->header_ops || !dev->header_ops->parse) 2696 return 0; 2697 return dev->header_ops->parse(skb, haddr); 2698 } 2699 2700 /* ll_header must have at least hard_header_len allocated */ 2701 static inline bool dev_validate_header(const struct net_device *dev, 2702 char *ll_header, int len) 2703 { 2704 if (likely(len >= dev->hard_header_len)) 2705 return true; 2706 if (len < dev->min_header_len) 2707 return false; 2708 2709 if (capable(CAP_SYS_RAWIO)) { 2710 memset(ll_header + len, 0, dev->hard_header_len - len); 2711 return true; 2712 } 2713 2714 if (dev->header_ops && dev->header_ops->validate) 2715 return dev->header_ops->validate(ll_header, len); 2716 2717 return false; 2718 } 2719 2720 typedef int gifconf_func_t(struct net_device * dev, char __user * bufptr, int len); 2721 int register_gifconf(unsigned int family, gifconf_func_t *gifconf); 2722 static inline int unregister_gifconf(unsigned int family) 2723 { 2724 return register_gifconf(family, NULL); 2725 } 2726 2727 #ifdef CONFIG_NET_FLOW_LIMIT 2728 #define FLOW_LIMIT_HISTORY (1 << 7) /* must be ^2 and !overflow buckets */ 2729 struct sd_flow_limit { 2730 u64 count; 2731 unsigned int num_buckets; 2732 unsigned int history_head; 2733 u16 history[FLOW_LIMIT_HISTORY]; 2734 u8 buckets[]; 2735 }; 2736 2737 extern int netdev_flow_limit_table_len; 2738 #endif /* CONFIG_NET_FLOW_LIMIT */ 2739 2740 /* 2741 * Incoming packets are placed on per-CPU queues 2742 */ 2743 struct softnet_data { 2744 struct list_head poll_list; 2745 struct sk_buff_head process_queue; 2746 2747 /* stats */ 2748 unsigned int processed; 2749 unsigned int time_squeeze; 2750 unsigned int received_rps; 2751 #ifdef CONFIG_RPS 2752 struct softnet_data *rps_ipi_list; 2753 #endif 2754 #ifdef CONFIG_NET_FLOW_LIMIT 2755 struct sd_flow_limit __rcu *flow_limit; 2756 #endif 2757 struct Qdisc *output_queue; 2758 struct Qdisc **output_queue_tailp; 2759 struct sk_buff *completion_queue; 2760 2761 #ifdef CONFIG_RPS 2762 /* input_queue_head should be written by cpu owning this struct, 2763 * and only read by other cpus. Worth using a cache line. 2764 */ 2765 unsigned int input_queue_head ____cacheline_aligned_in_smp; 2766 2767 /* Elements below can be accessed between CPUs for RPS/RFS */ 2768 struct call_single_data csd ____cacheline_aligned_in_smp; 2769 struct softnet_data *rps_ipi_next; 2770 unsigned int cpu; 2771 unsigned int input_queue_tail; 2772 #endif 2773 unsigned int dropped; 2774 struct sk_buff_head input_pkt_queue; 2775 struct napi_struct backlog; 2776 2777 }; 2778 2779 static inline void input_queue_head_incr(struct softnet_data *sd) 2780 { 2781 #ifdef CONFIG_RPS 2782 sd->input_queue_head++; 2783 #endif 2784 } 2785 2786 static inline void input_queue_tail_incr_save(struct softnet_data *sd, 2787 unsigned int *qtail) 2788 { 2789 #ifdef CONFIG_RPS 2790 *qtail = ++sd->input_queue_tail; 2791 #endif 2792 } 2793 2794 DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data); 2795 2796 void __netif_schedule(struct Qdisc *q); 2797 void netif_schedule_queue(struct netdev_queue *txq); 2798 2799 static inline void netif_tx_schedule_all(struct net_device *dev) 2800 { 2801 unsigned int i; 2802 2803 for (i = 0; i < dev->num_tx_queues; i++) 2804 netif_schedule_queue(netdev_get_tx_queue(dev, i)); 2805 } 2806 2807 static __always_inline void netif_tx_start_queue(struct netdev_queue *dev_queue) 2808 { 2809 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state); 2810 } 2811 2812 /** 2813 * netif_start_queue - allow transmit 2814 * @dev: network device 2815 * 2816 * Allow upper layers to call the device hard_start_xmit routine. 2817 */ 2818 static inline void netif_start_queue(struct net_device *dev) 2819 { 2820 netif_tx_start_queue(netdev_get_tx_queue(dev, 0)); 2821 } 2822 2823 static inline void netif_tx_start_all_queues(struct net_device *dev) 2824 { 2825 unsigned int i; 2826 2827 for (i = 0; i < dev->num_tx_queues; i++) { 2828 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 2829 netif_tx_start_queue(txq); 2830 } 2831 } 2832 2833 void netif_tx_wake_queue(struct netdev_queue *dev_queue); 2834 2835 /** 2836 * netif_wake_queue - restart transmit 2837 * @dev: network device 2838 * 2839 * Allow upper layers to call the device hard_start_xmit routine. 2840 * Used for flow control when transmit resources are available. 2841 */ 2842 static inline void netif_wake_queue(struct net_device *dev) 2843 { 2844 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0)); 2845 } 2846 2847 static inline void netif_tx_wake_all_queues(struct net_device *dev) 2848 { 2849 unsigned int i; 2850 2851 for (i = 0; i < dev->num_tx_queues; i++) { 2852 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 2853 netif_tx_wake_queue(txq); 2854 } 2855 } 2856 2857 static __always_inline void netif_tx_stop_queue(struct netdev_queue *dev_queue) 2858 { 2859 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state); 2860 } 2861 2862 /** 2863 * netif_stop_queue - stop transmitted packets 2864 * @dev: network device 2865 * 2866 * Stop upper layers calling the device hard_start_xmit routine. 2867 * Used for flow control when transmit resources are unavailable. 2868 */ 2869 static inline void netif_stop_queue(struct net_device *dev) 2870 { 2871 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0)); 2872 } 2873 2874 void netif_tx_stop_all_queues(struct net_device *dev); 2875 2876 static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue) 2877 { 2878 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state); 2879 } 2880 2881 /** 2882 * netif_queue_stopped - test if transmit queue is flowblocked 2883 * @dev: network device 2884 * 2885 * Test if transmit queue on device is currently unable to send. 2886 */ 2887 static inline bool netif_queue_stopped(const struct net_device *dev) 2888 { 2889 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0)); 2890 } 2891 2892 static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue) 2893 { 2894 return dev_queue->state & QUEUE_STATE_ANY_XOFF; 2895 } 2896 2897 static inline bool 2898 netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue) 2899 { 2900 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN; 2901 } 2902 2903 static inline bool 2904 netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue) 2905 { 2906 return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN; 2907 } 2908 2909 /** 2910 * netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write 2911 * @dev_queue: pointer to transmit queue 2912 * 2913 * BQL enabled drivers might use this helper in their ndo_start_xmit(), 2914 * to give appropriate hint to the CPU. 2915 */ 2916 static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue) 2917 { 2918 #ifdef CONFIG_BQL 2919 prefetchw(&dev_queue->dql.num_queued); 2920 #endif 2921 } 2922 2923 /** 2924 * netdev_txq_bql_complete_prefetchw - prefetch bql data for write 2925 * @dev_queue: pointer to transmit queue 2926 * 2927 * BQL enabled drivers might use this helper in their TX completion path, 2928 * to give appropriate hint to the CPU. 2929 */ 2930 static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue) 2931 { 2932 #ifdef CONFIG_BQL 2933 prefetchw(&dev_queue->dql.limit); 2934 #endif 2935 } 2936 2937 static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue, 2938 unsigned int bytes) 2939 { 2940 #ifdef CONFIG_BQL 2941 dql_queued(&dev_queue->dql, bytes); 2942 2943 if (likely(dql_avail(&dev_queue->dql) >= 0)) 2944 return; 2945 2946 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state); 2947 2948 /* 2949 * The XOFF flag must be set before checking the dql_avail below, 2950 * because in netdev_tx_completed_queue we update the dql_completed 2951 * before checking the XOFF flag. 2952 */ 2953 smp_mb(); 2954 2955 /* check again in case another CPU has just made room avail */ 2956 if (unlikely(dql_avail(&dev_queue->dql) >= 0)) 2957 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state); 2958 #endif 2959 } 2960 2961 /** 2962 * netdev_sent_queue - report the number of bytes queued to hardware 2963 * @dev: network device 2964 * @bytes: number of bytes queued to the hardware device queue 2965 * 2966 * Report the number of bytes queued for sending/completion to the network 2967 * device hardware queue. @bytes should be a good approximation and should 2968 * exactly match netdev_completed_queue() @bytes 2969 */ 2970 static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes) 2971 { 2972 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes); 2973 } 2974 2975 static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue, 2976 unsigned int pkts, unsigned int bytes) 2977 { 2978 #ifdef CONFIG_BQL 2979 if (unlikely(!bytes)) 2980 return; 2981 2982 dql_completed(&dev_queue->dql, bytes); 2983 2984 /* 2985 * Without the memory barrier there is a small possiblity that 2986 * netdev_tx_sent_queue will miss the update and cause the queue to 2987 * be stopped forever 2988 */ 2989 smp_mb(); 2990 2991 if (dql_avail(&dev_queue->dql) < 0) 2992 return; 2993 2994 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state)) 2995 netif_schedule_queue(dev_queue); 2996 #endif 2997 } 2998 2999 /** 3000 * netdev_completed_queue - report bytes and packets completed by device 3001 * @dev: network device 3002 * @pkts: actual number of packets sent over the medium 3003 * @bytes: actual number of bytes sent over the medium 3004 * 3005 * Report the number of bytes and packets transmitted by the network device 3006 * hardware queue over the physical medium, @bytes must exactly match the 3007 * @bytes amount passed to netdev_sent_queue() 3008 */ 3009 static inline void netdev_completed_queue(struct net_device *dev, 3010 unsigned int pkts, unsigned int bytes) 3011 { 3012 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes); 3013 } 3014 3015 static inline void netdev_tx_reset_queue(struct netdev_queue *q) 3016 { 3017 #ifdef CONFIG_BQL 3018 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state); 3019 dql_reset(&q->dql); 3020 #endif 3021 } 3022 3023 /** 3024 * netdev_reset_queue - reset the packets and bytes count of a network device 3025 * @dev_queue: network device 3026 * 3027 * Reset the bytes and packet count of a network device and clear the 3028 * software flow control OFF bit for this network device 3029 */ 3030 static inline void netdev_reset_queue(struct net_device *dev_queue) 3031 { 3032 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0)); 3033 } 3034 3035 /** 3036 * netdev_cap_txqueue - check if selected tx queue exceeds device queues 3037 * @dev: network device 3038 * @queue_index: given tx queue index 3039 * 3040 * Returns 0 if given tx queue index >= number of device tx queues, 3041 * otherwise returns the originally passed tx queue index. 3042 */ 3043 static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index) 3044 { 3045 if (unlikely(queue_index >= dev->real_num_tx_queues)) { 3046 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n", 3047 dev->name, queue_index, 3048 dev->real_num_tx_queues); 3049 return 0; 3050 } 3051 3052 return queue_index; 3053 } 3054 3055 /** 3056 * netif_running - test if up 3057 * @dev: network device 3058 * 3059 * Test if the device has been brought up. 3060 */ 3061 static inline bool netif_running(const struct net_device *dev) 3062 { 3063 return test_bit(__LINK_STATE_START, &dev->state); 3064 } 3065 3066 /* 3067 * Routines to manage the subqueues on a device. We only need start, 3068 * stop, and a check if it's stopped. All other device management is 3069 * done at the overall netdevice level. 3070 * Also test the device if we're multiqueue. 3071 */ 3072 3073 /** 3074 * netif_start_subqueue - allow sending packets on subqueue 3075 * @dev: network device 3076 * @queue_index: sub queue index 3077 * 3078 * Start individual transmit queue of a device with multiple transmit queues. 3079 */ 3080 static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index) 3081 { 3082 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 3083 3084 netif_tx_start_queue(txq); 3085 } 3086 3087 /** 3088 * netif_stop_subqueue - stop sending packets on subqueue 3089 * @dev: network device 3090 * @queue_index: sub queue index 3091 * 3092 * Stop individual transmit queue of a device with multiple transmit queues. 3093 */ 3094 static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index) 3095 { 3096 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 3097 netif_tx_stop_queue(txq); 3098 } 3099 3100 /** 3101 * netif_subqueue_stopped - test status of subqueue 3102 * @dev: network device 3103 * @queue_index: sub queue index 3104 * 3105 * Check individual transmit queue of a device with multiple transmit queues. 3106 */ 3107 static inline bool __netif_subqueue_stopped(const struct net_device *dev, 3108 u16 queue_index) 3109 { 3110 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 3111 3112 return netif_tx_queue_stopped(txq); 3113 } 3114 3115 static inline bool netif_subqueue_stopped(const struct net_device *dev, 3116 struct sk_buff *skb) 3117 { 3118 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb)); 3119 } 3120 3121 /** 3122 * netif_wake_subqueue - allow sending packets on subqueue 3123 * @dev: network device 3124 * @queue_index: sub queue index 3125 * 3126 * Resume individual transmit queue of a device with multiple transmit queues. 3127 */ 3128 static inline void netif_wake_subqueue(struct net_device *dev, u16 queue_index) 3129 { 3130 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 3131 3132 netif_tx_wake_queue(txq); 3133 } 3134 3135 #ifdef CONFIG_XPS 3136 int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask, 3137 u16 index); 3138 #else 3139 static inline int netif_set_xps_queue(struct net_device *dev, 3140 const struct cpumask *mask, 3141 u16 index) 3142 { 3143 return 0; 3144 } 3145 #endif 3146 3147 u16 __skb_tx_hash(const struct net_device *dev, struct sk_buff *skb, 3148 unsigned int num_tx_queues); 3149 3150 /* 3151 * Returns a Tx hash for the given packet when dev->real_num_tx_queues is used 3152 * as a distribution range limit for the returned value. 3153 */ 3154 static inline u16 skb_tx_hash(const struct net_device *dev, 3155 struct sk_buff *skb) 3156 { 3157 return __skb_tx_hash(dev, skb, dev->real_num_tx_queues); 3158 } 3159 3160 /** 3161 * netif_is_multiqueue - test if device has multiple transmit queues 3162 * @dev: network device 3163 * 3164 * Check if device has multiple transmit queues 3165 */ 3166 static inline bool netif_is_multiqueue(const struct net_device *dev) 3167 { 3168 return dev->num_tx_queues > 1; 3169 } 3170 3171 int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq); 3172 3173 #ifdef CONFIG_SYSFS 3174 int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq); 3175 #else 3176 static inline int netif_set_real_num_rx_queues(struct net_device *dev, 3177 unsigned int rxq) 3178 { 3179 return 0; 3180 } 3181 #endif 3182 3183 #ifdef CONFIG_SYSFS 3184 static inline unsigned int get_netdev_rx_queue_index( 3185 struct netdev_rx_queue *queue) 3186 { 3187 struct net_device *dev = queue->dev; 3188 int index = queue - dev->_rx; 3189 3190 BUG_ON(index >= dev->num_rx_queues); 3191 return index; 3192 } 3193 #endif 3194 3195 #define DEFAULT_MAX_NUM_RSS_QUEUES (8) 3196 int netif_get_num_default_rss_queues(void); 3197 3198 enum skb_free_reason { 3199 SKB_REASON_CONSUMED, 3200 SKB_REASON_DROPPED, 3201 }; 3202 3203 void __dev_kfree_skb_irq(struct sk_buff *skb, enum skb_free_reason reason); 3204 void __dev_kfree_skb_any(struct sk_buff *skb, enum skb_free_reason reason); 3205 3206 /* 3207 * It is not allowed to call kfree_skb() or consume_skb() from hardware 3208 * interrupt context or with hardware interrupts being disabled. 3209 * (in_irq() || irqs_disabled()) 3210 * 3211 * We provide four helpers that can be used in following contexts : 3212 * 3213 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context, 3214 * replacing kfree_skb(skb) 3215 * 3216 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context. 3217 * Typically used in place of consume_skb(skb) in TX completion path 3218 * 3219 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context, 3220 * replacing kfree_skb(skb) 3221 * 3222 * dev_consume_skb_any(skb) when caller doesn't know its current irq context, 3223 * and consumed a packet. Used in place of consume_skb(skb) 3224 */ 3225 static inline void dev_kfree_skb_irq(struct sk_buff *skb) 3226 { 3227 __dev_kfree_skb_irq(skb, SKB_REASON_DROPPED); 3228 } 3229 3230 static inline void dev_consume_skb_irq(struct sk_buff *skb) 3231 { 3232 __dev_kfree_skb_irq(skb, SKB_REASON_CONSUMED); 3233 } 3234 3235 static inline void dev_kfree_skb_any(struct sk_buff *skb) 3236 { 3237 __dev_kfree_skb_any(skb, SKB_REASON_DROPPED); 3238 } 3239 3240 static inline void dev_consume_skb_any(struct sk_buff *skb) 3241 { 3242 __dev_kfree_skb_any(skb, SKB_REASON_CONSUMED); 3243 } 3244 3245 void generic_xdp_tx(struct sk_buff *skb, struct bpf_prog *xdp_prog); 3246 int do_xdp_generic(struct bpf_prog *xdp_prog, struct sk_buff *skb); 3247 int netif_rx(struct sk_buff *skb); 3248 int netif_rx_ni(struct sk_buff *skb); 3249 int netif_receive_skb(struct sk_buff *skb); 3250 gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb); 3251 void napi_gro_flush(struct napi_struct *napi, bool flush_old); 3252 struct sk_buff *napi_get_frags(struct napi_struct *napi); 3253 gro_result_t napi_gro_frags(struct napi_struct *napi); 3254 struct packet_offload *gro_find_receive_by_type(__be16 type); 3255 struct packet_offload *gro_find_complete_by_type(__be16 type); 3256 3257 static inline void napi_free_frags(struct napi_struct *napi) 3258 { 3259 kfree_skb(napi->skb); 3260 napi->skb = NULL; 3261 } 3262 3263 bool netdev_is_rx_handler_busy(struct net_device *dev); 3264 int netdev_rx_handler_register(struct net_device *dev, 3265 rx_handler_func_t *rx_handler, 3266 void *rx_handler_data); 3267 void netdev_rx_handler_unregister(struct net_device *dev); 3268 3269 bool dev_valid_name(const char *name); 3270 int dev_ioctl(struct net *net, unsigned int cmd, void __user *); 3271 int dev_ethtool(struct net *net, struct ifreq *); 3272 unsigned int dev_get_flags(const struct net_device *); 3273 int __dev_change_flags(struct net_device *, unsigned int flags); 3274 int dev_change_flags(struct net_device *, unsigned int); 3275 void __dev_notify_flags(struct net_device *, unsigned int old_flags, 3276 unsigned int gchanges); 3277 int dev_change_name(struct net_device *, const char *); 3278 int dev_set_alias(struct net_device *, const char *, size_t); 3279 int dev_change_net_namespace(struct net_device *, struct net *, const char *); 3280 int __dev_set_mtu(struct net_device *, int); 3281 int dev_set_mtu(struct net_device *, int); 3282 void dev_set_group(struct net_device *, int); 3283 int dev_set_mac_address(struct net_device *, struct sockaddr *); 3284 int dev_change_carrier(struct net_device *, bool new_carrier); 3285 int dev_get_phys_port_id(struct net_device *dev, 3286 struct netdev_phys_item_id *ppid); 3287 int dev_get_phys_port_name(struct net_device *dev, 3288 char *name, size_t len); 3289 int dev_change_proto_down(struct net_device *dev, bool proto_down); 3290 struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev); 3291 struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev, 3292 struct netdev_queue *txq, int *ret); 3293 3294 typedef int (*xdp_op_t)(struct net_device *dev, struct netdev_xdp *xdp); 3295 int dev_change_xdp_fd(struct net_device *dev, struct netlink_ext_ack *extack, 3296 int fd, u32 flags); 3297 u8 __dev_xdp_attached(struct net_device *dev, xdp_op_t xdp_op, u32 *prog_id); 3298 3299 int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb); 3300 int dev_forward_skb(struct net_device *dev, struct sk_buff *skb); 3301 bool is_skb_forwardable(const struct net_device *dev, 3302 const struct sk_buff *skb); 3303 3304 static __always_inline int ____dev_forward_skb(struct net_device *dev, 3305 struct sk_buff *skb) 3306 { 3307 if (skb_orphan_frags(skb, GFP_ATOMIC) || 3308 unlikely(!is_skb_forwardable(dev, skb))) { 3309 atomic_long_inc(&dev->rx_dropped); 3310 kfree_skb(skb); 3311 return NET_RX_DROP; 3312 } 3313 3314 skb_scrub_packet(skb, true); 3315 skb->priority = 0; 3316 return 0; 3317 } 3318 3319 void dev_queue_xmit_nit(struct sk_buff *skb, struct net_device *dev); 3320 3321 extern int netdev_budget; 3322 extern unsigned int netdev_budget_usecs; 3323 3324 /* Called by rtnetlink.c:rtnl_unlock() */ 3325 void netdev_run_todo(void); 3326 3327 /** 3328 * dev_put - release reference to device 3329 * @dev: network device 3330 * 3331 * Release reference to device to allow it to be freed. 3332 */ 3333 static inline void dev_put(struct net_device *dev) 3334 { 3335 this_cpu_dec(*dev->pcpu_refcnt); 3336 } 3337 3338 /** 3339 * dev_hold - get reference to device 3340 * @dev: network device 3341 * 3342 * Hold reference to device to keep it from being freed. 3343 */ 3344 static inline void dev_hold(struct net_device *dev) 3345 { 3346 this_cpu_inc(*dev->pcpu_refcnt); 3347 } 3348 3349 /* Carrier loss detection, dial on demand. The functions netif_carrier_on 3350 * and _off may be called from IRQ context, but it is caller 3351 * who is responsible for serialization of these calls. 3352 * 3353 * The name carrier is inappropriate, these functions should really be 3354 * called netif_lowerlayer_*() because they represent the state of any 3355 * kind of lower layer not just hardware media. 3356 */ 3357 3358 void linkwatch_init_dev(struct net_device *dev); 3359 void linkwatch_fire_event(struct net_device *dev); 3360 void linkwatch_forget_dev(struct net_device *dev); 3361 3362 /** 3363 * netif_carrier_ok - test if carrier present 3364 * @dev: network device 3365 * 3366 * Check if carrier is present on device 3367 */ 3368 static inline bool netif_carrier_ok(const struct net_device *dev) 3369 { 3370 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state); 3371 } 3372 3373 unsigned long dev_trans_start(struct net_device *dev); 3374 3375 void __netdev_watchdog_up(struct net_device *dev); 3376 3377 void netif_carrier_on(struct net_device *dev); 3378 3379 void netif_carrier_off(struct net_device *dev); 3380 3381 /** 3382 * netif_dormant_on - mark device as dormant. 3383 * @dev: network device 3384 * 3385 * Mark device as dormant (as per RFC2863). 3386 * 3387 * The dormant state indicates that the relevant interface is not 3388 * actually in a condition to pass packets (i.e., it is not 'up') but is 3389 * in a "pending" state, waiting for some external event. For "on- 3390 * demand" interfaces, this new state identifies the situation where the 3391 * interface is waiting for events to place it in the up state. 3392 */ 3393 static inline void netif_dormant_on(struct net_device *dev) 3394 { 3395 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state)) 3396 linkwatch_fire_event(dev); 3397 } 3398 3399 /** 3400 * netif_dormant_off - set device as not dormant. 3401 * @dev: network device 3402 * 3403 * Device is not in dormant state. 3404 */ 3405 static inline void netif_dormant_off(struct net_device *dev) 3406 { 3407 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state)) 3408 linkwatch_fire_event(dev); 3409 } 3410 3411 /** 3412 * netif_dormant - test if device is dormant 3413 * @dev: network device 3414 * 3415 * Check if device is dormant. 3416 */ 3417 static inline bool netif_dormant(const struct net_device *dev) 3418 { 3419 return test_bit(__LINK_STATE_DORMANT, &dev->state); 3420 } 3421 3422 3423 /** 3424 * netif_oper_up - test if device is operational 3425 * @dev: network device 3426 * 3427 * Check if carrier is operational 3428 */ 3429 static inline bool netif_oper_up(const struct net_device *dev) 3430 { 3431 return (dev->operstate == IF_OPER_UP || 3432 dev->operstate == IF_OPER_UNKNOWN /* backward compat */); 3433 } 3434 3435 /** 3436 * netif_device_present - is device available or removed 3437 * @dev: network device 3438 * 3439 * Check if device has not been removed from system. 3440 */ 3441 static inline bool netif_device_present(struct net_device *dev) 3442 { 3443 return test_bit(__LINK_STATE_PRESENT, &dev->state); 3444 } 3445 3446 void netif_device_detach(struct net_device *dev); 3447 3448 void netif_device_attach(struct net_device *dev); 3449 3450 /* 3451 * Network interface message level settings 3452 */ 3453 3454 enum { 3455 NETIF_MSG_DRV = 0x0001, 3456 NETIF_MSG_PROBE = 0x0002, 3457 NETIF_MSG_LINK = 0x0004, 3458 NETIF_MSG_TIMER = 0x0008, 3459 NETIF_MSG_IFDOWN = 0x0010, 3460 NETIF_MSG_IFUP = 0x0020, 3461 NETIF_MSG_RX_ERR = 0x0040, 3462 NETIF_MSG_TX_ERR = 0x0080, 3463 NETIF_MSG_TX_QUEUED = 0x0100, 3464 NETIF_MSG_INTR = 0x0200, 3465 NETIF_MSG_TX_DONE = 0x0400, 3466 NETIF_MSG_RX_STATUS = 0x0800, 3467 NETIF_MSG_PKTDATA = 0x1000, 3468 NETIF_MSG_HW = 0x2000, 3469 NETIF_MSG_WOL = 0x4000, 3470 }; 3471 3472 #define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV) 3473 #define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE) 3474 #define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK) 3475 #define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER) 3476 #define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN) 3477 #define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP) 3478 #define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR) 3479 #define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR) 3480 #define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED) 3481 #define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR) 3482 #define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE) 3483 #define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS) 3484 #define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA) 3485 #define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW) 3486 #define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL) 3487 3488 static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits) 3489 { 3490 /* use default */ 3491 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8)) 3492 return default_msg_enable_bits; 3493 if (debug_value == 0) /* no output */ 3494 return 0; 3495 /* set low N bits */ 3496 return (1 << debug_value) - 1; 3497 } 3498 3499 static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu) 3500 { 3501 spin_lock(&txq->_xmit_lock); 3502 txq->xmit_lock_owner = cpu; 3503 } 3504 3505 static inline bool __netif_tx_acquire(struct netdev_queue *txq) 3506 { 3507 __acquire(&txq->_xmit_lock); 3508 return true; 3509 } 3510 3511 static inline void __netif_tx_release(struct netdev_queue *txq) 3512 { 3513 __release(&txq->_xmit_lock); 3514 } 3515 3516 static inline void __netif_tx_lock_bh(struct netdev_queue *txq) 3517 { 3518 spin_lock_bh(&txq->_xmit_lock); 3519 txq->xmit_lock_owner = smp_processor_id(); 3520 } 3521 3522 static inline bool __netif_tx_trylock(struct netdev_queue *txq) 3523 { 3524 bool ok = spin_trylock(&txq->_xmit_lock); 3525 if (likely(ok)) 3526 txq->xmit_lock_owner = smp_processor_id(); 3527 return ok; 3528 } 3529 3530 static inline void __netif_tx_unlock(struct netdev_queue *txq) 3531 { 3532 txq->xmit_lock_owner = -1; 3533 spin_unlock(&txq->_xmit_lock); 3534 } 3535 3536 static inline void __netif_tx_unlock_bh(struct netdev_queue *txq) 3537 { 3538 txq->xmit_lock_owner = -1; 3539 spin_unlock_bh(&txq->_xmit_lock); 3540 } 3541 3542 static inline void txq_trans_update(struct netdev_queue *txq) 3543 { 3544 if (txq->xmit_lock_owner != -1) 3545 txq->trans_start = jiffies; 3546 } 3547 3548 /* legacy drivers only, netdev_start_xmit() sets txq->trans_start */ 3549 static inline void netif_trans_update(struct net_device *dev) 3550 { 3551 struct netdev_queue *txq = netdev_get_tx_queue(dev, 0); 3552 3553 if (txq->trans_start != jiffies) 3554 txq->trans_start = jiffies; 3555 } 3556 3557 /** 3558 * netif_tx_lock - grab network device transmit lock 3559 * @dev: network device 3560 * 3561 * Get network device transmit lock 3562 */ 3563 static inline void netif_tx_lock(struct net_device *dev) 3564 { 3565 unsigned int i; 3566 int cpu; 3567 3568 spin_lock(&dev->tx_global_lock); 3569 cpu = smp_processor_id(); 3570 for (i = 0; i < dev->num_tx_queues; i++) { 3571 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 3572 3573 /* We are the only thread of execution doing a 3574 * freeze, but we have to grab the _xmit_lock in 3575 * order to synchronize with threads which are in 3576 * the ->hard_start_xmit() handler and already 3577 * checked the frozen bit. 3578 */ 3579 __netif_tx_lock(txq, cpu); 3580 set_bit(__QUEUE_STATE_FROZEN, &txq->state); 3581 __netif_tx_unlock(txq); 3582 } 3583 } 3584 3585 static inline void netif_tx_lock_bh(struct net_device *dev) 3586 { 3587 local_bh_disable(); 3588 netif_tx_lock(dev); 3589 } 3590 3591 static inline void netif_tx_unlock(struct net_device *dev) 3592 { 3593 unsigned int i; 3594 3595 for (i = 0; i < dev->num_tx_queues; i++) { 3596 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 3597 3598 /* No need to grab the _xmit_lock here. If the 3599 * queue is not stopped for another reason, we 3600 * force a schedule. 3601 */ 3602 clear_bit(__QUEUE_STATE_FROZEN, &txq->state); 3603 netif_schedule_queue(txq); 3604 } 3605 spin_unlock(&dev->tx_global_lock); 3606 } 3607 3608 static inline void netif_tx_unlock_bh(struct net_device *dev) 3609 { 3610 netif_tx_unlock(dev); 3611 local_bh_enable(); 3612 } 3613 3614 #define HARD_TX_LOCK(dev, txq, cpu) { \ 3615 if ((dev->features & NETIF_F_LLTX) == 0) { \ 3616 __netif_tx_lock(txq, cpu); \ 3617 } else { \ 3618 __netif_tx_acquire(txq); \ 3619 } \ 3620 } 3621 3622 #define HARD_TX_TRYLOCK(dev, txq) \ 3623 (((dev->features & NETIF_F_LLTX) == 0) ? \ 3624 __netif_tx_trylock(txq) : \ 3625 __netif_tx_acquire(txq)) 3626 3627 #define HARD_TX_UNLOCK(dev, txq) { \ 3628 if ((dev->features & NETIF_F_LLTX) == 0) { \ 3629 __netif_tx_unlock(txq); \ 3630 } else { \ 3631 __netif_tx_release(txq); \ 3632 } \ 3633 } 3634 3635 static inline void netif_tx_disable(struct net_device *dev) 3636 { 3637 unsigned int i; 3638 int cpu; 3639 3640 local_bh_disable(); 3641 cpu = smp_processor_id(); 3642 for (i = 0; i < dev->num_tx_queues; i++) { 3643 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 3644 3645 __netif_tx_lock(txq, cpu); 3646 netif_tx_stop_queue(txq); 3647 __netif_tx_unlock(txq); 3648 } 3649 local_bh_enable(); 3650 } 3651 3652 static inline void netif_addr_lock(struct net_device *dev) 3653 { 3654 spin_lock(&dev->addr_list_lock); 3655 } 3656 3657 static inline void netif_addr_lock_nested(struct net_device *dev) 3658 { 3659 int subclass = SINGLE_DEPTH_NESTING; 3660 3661 if (dev->netdev_ops->ndo_get_lock_subclass) 3662 subclass = dev->netdev_ops->ndo_get_lock_subclass(dev); 3663 3664 spin_lock_nested(&dev->addr_list_lock, subclass); 3665 } 3666 3667 static inline void netif_addr_lock_bh(struct net_device *dev) 3668 { 3669 spin_lock_bh(&dev->addr_list_lock); 3670 } 3671 3672 static inline void netif_addr_unlock(struct net_device *dev) 3673 { 3674 spin_unlock(&dev->addr_list_lock); 3675 } 3676 3677 static inline void netif_addr_unlock_bh(struct net_device *dev) 3678 { 3679 spin_unlock_bh(&dev->addr_list_lock); 3680 } 3681 3682 /* 3683 * dev_addrs walker. Should be used only for read access. Call with 3684 * rcu_read_lock held. 3685 */ 3686 #define for_each_dev_addr(dev, ha) \ 3687 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list) 3688 3689 /* These functions live elsewhere (drivers/net/net_init.c, but related) */ 3690 3691 void ether_setup(struct net_device *dev); 3692 3693 /* Support for loadable net-drivers */ 3694 struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name, 3695 unsigned char name_assign_type, 3696 void (*setup)(struct net_device *), 3697 unsigned int txqs, unsigned int rxqs); 3698 #define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \ 3699 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1) 3700 3701 #define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \ 3702 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \ 3703 count) 3704 3705 int register_netdev(struct net_device *dev); 3706 void unregister_netdev(struct net_device *dev); 3707 3708 /* General hardware address lists handling functions */ 3709 int __hw_addr_sync(struct netdev_hw_addr_list *to_list, 3710 struct netdev_hw_addr_list *from_list, int addr_len); 3711 void __hw_addr_unsync(struct netdev_hw_addr_list *to_list, 3712 struct netdev_hw_addr_list *from_list, int addr_len); 3713 int __hw_addr_sync_dev(struct netdev_hw_addr_list *list, 3714 struct net_device *dev, 3715 int (*sync)(struct net_device *, const unsigned char *), 3716 int (*unsync)(struct net_device *, 3717 const unsigned char *)); 3718 void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list, 3719 struct net_device *dev, 3720 int (*unsync)(struct net_device *, 3721 const unsigned char *)); 3722 void __hw_addr_init(struct netdev_hw_addr_list *list); 3723 3724 /* Functions used for device addresses handling */ 3725 int dev_addr_add(struct net_device *dev, const unsigned char *addr, 3726 unsigned char addr_type); 3727 int dev_addr_del(struct net_device *dev, const unsigned char *addr, 3728 unsigned char addr_type); 3729 void dev_addr_flush(struct net_device *dev); 3730 int dev_addr_init(struct net_device *dev); 3731 3732 /* Functions used for unicast addresses handling */ 3733 int dev_uc_add(struct net_device *dev, const unsigned char *addr); 3734 int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr); 3735 int dev_uc_del(struct net_device *dev, const unsigned char *addr); 3736 int dev_uc_sync(struct net_device *to, struct net_device *from); 3737 int dev_uc_sync_multiple(struct net_device *to, struct net_device *from); 3738 void dev_uc_unsync(struct net_device *to, struct net_device *from); 3739 void dev_uc_flush(struct net_device *dev); 3740 void dev_uc_init(struct net_device *dev); 3741 3742 /** 3743 * __dev_uc_sync - Synchonize device's unicast list 3744 * @dev: device to sync 3745 * @sync: function to call if address should be added 3746 * @unsync: function to call if address should be removed 3747 * 3748 * Add newly added addresses to the interface, and release 3749 * addresses that have been deleted. 3750 */ 3751 static inline int __dev_uc_sync(struct net_device *dev, 3752 int (*sync)(struct net_device *, 3753 const unsigned char *), 3754 int (*unsync)(struct net_device *, 3755 const unsigned char *)) 3756 { 3757 return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync); 3758 } 3759 3760 /** 3761 * __dev_uc_unsync - Remove synchronized addresses from device 3762 * @dev: device to sync 3763 * @unsync: function to call if address should be removed 3764 * 3765 * Remove all addresses that were added to the device by dev_uc_sync(). 3766 */ 3767 static inline void __dev_uc_unsync(struct net_device *dev, 3768 int (*unsync)(struct net_device *, 3769 const unsigned char *)) 3770 { 3771 __hw_addr_unsync_dev(&dev->uc, dev, unsync); 3772 } 3773 3774 /* Functions used for multicast addresses handling */ 3775 int dev_mc_add(struct net_device *dev, const unsigned char *addr); 3776 int dev_mc_add_global(struct net_device *dev, const unsigned char *addr); 3777 int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr); 3778 int dev_mc_del(struct net_device *dev, const unsigned char *addr); 3779 int dev_mc_del_global(struct net_device *dev, const unsigned char *addr); 3780 int dev_mc_sync(struct net_device *to, struct net_device *from); 3781 int dev_mc_sync_multiple(struct net_device *to, struct net_device *from); 3782 void dev_mc_unsync(struct net_device *to, struct net_device *from); 3783 void dev_mc_flush(struct net_device *dev); 3784 void dev_mc_init(struct net_device *dev); 3785 3786 /** 3787 * __dev_mc_sync - Synchonize device's multicast list 3788 * @dev: device to sync 3789 * @sync: function to call if address should be added 3790 * @unsync: function to call if address should be removed 3791 * 3792 * Add newly added addresses to the interface, and release 3793 * addresses that have been deleted. 3794 */ 3795 static inline int __dev_mc_sync(struct net_device *dev, 3796 int (*sync)(struct net_device *, 3797 const unsigned char *), 3798 int (*unsync)(struct net_device *, 3799 const unsigned char *)) 3800 { 3801 return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync); 3802 } 3803 3804 /** 3805 * __dev_mc_unsync - Remove synchronized addresses from device 3806 * @dev: device to sync 3807 * @unsync: function to call if address should be removed 3808 * 3809 * Remove all addresses that were added to the device by dev_mc_sync(). 3810 */ 3811 static inline void __dev_mc_unsync(struct net_device *dev, 3812 int (*unsync)(struct net_device *, 3813 const unsigned char *)) 3814 { 3815 __hw_addr_unsync_dev(&dev->mc, dev, unsync); 3816 } 3817 3818 /* Functions used for secondary unicast and multicast support */ 3819 void dev_set_rx_mode(struct net_device *dev); 3820 void __dev_set_rx_mode(struct net_device *dev); 3821 int dev_set_promiscuity(struct net_device *dev, int inc); 3822 int dev_set_allmulti(struct net_device *dev, int inc); 3823 void netdev_state_change(struct net_device *dev); 3824 void netdev_notify_peers(struct net_device *dev); 3825 void netdev_features_change(struct net_device *dev); 3826 /* Load a device via the kmod */ 3827 void dev_load(struct net *net, const char *name); 3828 struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev, 3829 struct rtnl_link_stats64 *storage); 3830 void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64, 3831 const struct net_device_stats *netdev_stats); 3832 3833 extern int netdev_max_backlog; 3834 extern int netdev_tstamp_prequeue; 3835 extern int weight_p; 3836 extern int dev_weight_rx_bias; 3837 extern int dev_weight_tx_bias; 3838 extern int dev_rx_weight; 3839 extern int dev_tx_weight; 3840 3841 bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev); 3842 struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev, 3843 struct list_head **iter); 3844 struct net_device *netdev_all_upper_get_next_dev_rcu(struct net_device *dev, 3845 struct list_head **iter); 3846 3847 /* iterate through upper list, must be called under RCU read lock */ 3848 #define netdev_for_each_upper_dev_rcu(dev, updev, iter) \ 3849 for (iter = &(dev)->adj_list.upper, \ 3850 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \ 3851 updev; \ 3852 updev = netdev_upper_get_next_dev_rcu(dev, &(iter))) 3853 3854 int netdev_walk_all_upper_dev_rcu(struct net_device *dev, 3855 int (*fn)(struct net_device *upper_dev, 3856 void *data), 3857 void *data); 3858 3859 bool netdev_has_upper_dev_all_rcu(struct net_device *dev, 3860 struct net_device *upper_dev); 3861 3862 void *netdev_lower_get_next_private(struct net_device *dev, 3863 struct list_head **iter); 3864 void *netdev_lower_get_next_private_rcu(struct net_device *dev, 3865 struct list_head **iter); 3866 3867 #define netdev_for_each_lower_private(dev, priv, iter) \ 3868 for (iter = (dev)->adj_list.lower.next, \ 3869 priv = netdev_lower_get_next_private(dev, &(iter)); \ 3870 priv; \ 3871 priv = netdev_lower_get_next_private(dev, &(iter))) 3872 3873 #define netdev_for_each_lower_private_rcu(dev, priv, iter) \ 3874 for (iter = &(dev)->adj_list.lower, \ 3875 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \ 3876 priv; \ 3877 priv = netdev_lower_get_next_private_rcu(dev, &(iter))) 3878 3879 void *netdev_lower_get_next(struct net_device *dev, 3880 struct list_head **iter); 3881 3882 #define netdev_for_each_lower_dev(dev, ldev, iter) \ 3883 for (iter = (dev)->adj_list.lower.next, \ 3884 ldev = netdev_lower_get_next(dev, &(iter)); \ 3885 ldev; \ 3886 ldev = netdev_lower_get_next(dev, &(iter))) 3887 3888 struct net_device *netdev_all_lower_get_next(struct net_device *dev, 3889 struct list_head **iter); 3890 struct net_device *netdev_all_lower_get_next_rcu(struct net_device *dev, 3891 struct list_head **iter); 3892 3893 int netdev_walk_all_lower_dev(struct net_device *dev, 3894 int (*fn)(struct net_device *lower_dev, 3895 void *data), 3896 void *data); 3897 int netdev_walk_all_lower_dev_rcu(struct net_device *dev, 3898 int (*fn)(struct net_device *lower_dev, 3899 void *data), 3900 void *data); 3901 3902 void *netdev_adjacent_get_private(struct list_head *adj_list); 3903 void *netdev_lower_get_first_private_rcu(struct net_device *dev); 3904 struct net_device *netdev_master_upper_dev_get(struct net_device *dev); 3905 struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev); 3906 int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev); 3907 int netdev_master_upper_dev_link(struct net_device *dev, 3908 struct net_device *upper_dev, 3909 void *upper_priv, void *upper_info); 3910 void netdev_upper_dev_unlink(struct net_device *dev, 3911 struct net_device *upper_dev); 3912 void netdev_adjacent_rename_links(struct net_device *dev, char *oldname); 3913 void *netdev_lower_dev_get_private(struct net_device *dev, 3914 struct net_device *lower_dev); 3915 void netdev_lower_state_changed(struct net_device *lower_dev, 3916 void *lower_state_info); 3917 3918 /* RSS keys are 40 or 52 bytes long */ 3919 #define NETDEV_RSS_KEY_LEN 52 3920 extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN] __read_mostly; 3921 void netdev_rss_key_fill(void *buffer, size_t len); 3922 3923 int dev_get_nest_level(struct net_device *dev); 3924 int skb_checksum_help(struct sk_buff *skb); 3925 int skb_crc32c_csum_help(struct sk_buff *skb); 3926 int skb_csum_hwoffload_help(struct sk_buff *skb, 3927 const netdev_features_t features); 3928 3929 struct sk_buff *__skb_gso_segment(struct sk_buff *skb, 3930 netdev_features_t features, bool tx_path); 3931 struct sk_buff *skb_mac_gso_segment(struct sk_buff *skb, 3932 netdev_features_t features); 3933 3934 struct netdev_bonding_info { 3935 ifslave slave; 3936 ifbond master; 3937 }; 3938 3939 struct netdev_notifier_bonding_info { 3940 struct netdev_notifier_info info; /* must be first */ 3941 struct netdev_bonding_info bonding_info; 3942 }; 3943 3944 void netdev_bonding_info_change(struct net_device *dev, 3945 struct netdev_bonding_info *bonding_info); 3946 3947 static inline 3948 struct sk_buff *skb_gso_segment(struct sk_buff *skb, netdev_features_t features) 3949 { 3950 return __skb_gso_segment(skb, features, true); 3951 } 3952 __be16 skb_network_protocol(struct sk_buff *skb, int *depth); 3953 3954 static inline bool can_checksum_protocol(netdev_features_t features, 3955 __be16 protocol) 3956 { 3957 if (protocol == htons(ETH_P_FCOE)) 3958 return !!(features & NETIF_F_FCOE_CRC); 3959 3960 /* Assume this is an IP checksum (not SCTP CRC) */ 3961 3962 if (features & NETIF_F_HW_CSUM) { 3963 /* Can checksum everything */ 3964 return true; 3965 } 3966 3967 switch (protocol) { 3968 case htons(ETH_P_IP): 3969 return !!(features & NETIF_F_IP_CSUM); 3970 case htons(ETH_P_IPV6): 3971 return !!(features & NETIF_F_IPV6_CSUM); 3972 default: 3973 return false; 3974 } 3975 } 3976 3977 #ifdef CONFIG_BUG 3978 void netdev_rx_csum_fault(struct net_device *dev); 3979 #else 3980 static inline void netdev_rx_csum_fault(struct net_device *dev) 3981 { 3982 } 3983 #endif 3984 /* rx skb timestamps */ 3985 void net_enable_timestamp(void); 3986 void net_disable_timestamp(void); 3987 3988 #ifdef CONFIG_PROC_FS 3989 int __init dev_proc_init(void); 3990 #else 3991 #define dev_proc_init() 0 3992 #endif 3993 3994 static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops, 3995 struct sk_buff *skb, struct net_device *dev, 3996 bool more) 3997 { 3998 skb->xmit_more = more ? 1 : 0; 3999 return ops->ndo_start_xmit(skb, dev); 4000 } 4001 4002 static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev, 4003 struct netdev_queue *txq, bool more) 4004 { 4005 const struct net_device_ops *ops = dev->netdev_ops; 4006 int rc; 4007 4008 rc = __netdev_start_xmit(ops, skb, dev, more); 4009 if (rc == NETDEV_TX_OK) 4010 txq_trans_update(txq); 4011 4012 return rc; 4013 } 4014 4015 int netdev_class_create_file_ns(const struct class_attribute *class_attr, 4016 const void *ns); 4017 void netdev_class_remove_file_ns(const struct class_attribute *class_attr, 4018 const void *ns); 4019 4020 static inline int netdev_class_create_file(const struct class_attribute *class_attr) 4021 { 4022 return netdev_class_create_file_ns(class_attr, NULL); 4023 } 4024 4025 static inline void netdev_class_remove_file(const struct class_attribute *class_attr) 4026 { 4027 netdev_class_remove_file_ns(class_attr, NULL); 4028 } 4029 4030 extern const struct kobj_ns_type_operations net_ns_type_operations; 4031 4032 const char *netdev_drivername(const struct net_device *dev); 4033 4034 void linkwatch_run_queue(void); 4035 4036 static inline netdev_features_t netdev_intersect_features(netdev_features_t f1, 4037 netdev_features_t f2) 4038 { 4039 if ((f1 ^ f2) & NETIF_F_HW_CSUM) { 4040 if (f1 & NETIF_F_HW_CSUM) 4041 f1 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM); 4042 else 4043 f2 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM); 4044 } 4045 4046 return f1 & f2; 4047 } 4048 4049 static inline netdev_features_t netdev_get_wanted_features( 4050 struct net_device *dev) 4051 { 4052 return (dev->features & ~dev->hw_features) | dev->wanted_features; 4053 } 4054 netdev_features_t netdev_increment_features(netdev_features_t all, 4055 netdev_features_t one, netdev_features_t mask); 4056 4057 /* Allow TSO being used on stacked device : 4058 * Performing the GSO segmentation before last device 4059 * is a performance improvement. 4060 */ 4061 static inline netdev_features_t netdev_add_tso_features(netdev_features_t features, 4062 netdev_features_t mask) 4063 { 4064 return netdev_increment_features(features, NETIF_F_ALL_TSO, mask); 4065 } 4066 4067 int __netdev_update_features(struct net_device *dev); 4068 void netdev_update_features(struct net_device *dev); 4069 void netdev_change_features(struct net_device *dev); 4070 4071 void netif_stacked_transfer_operstate(const struct net_device *rootdev, 4072 struct net_device *dev); 4073 4074 netdev_features_t passthru_features_check(struct sk_buff *skb, 4075 struct net_device *dev, 4076 netdev_features_t features); 4077 netdev_features_t netif_skb_features(struct sk_buff *skb); 4078 4079 static inline bool net_gso_ok(netdev_features_t features, int gso_type) 4080 { 4081 netdev_features_t feature = (netdev_features_t)gso_type << NETIF_F_GSO_SHIFT; 4082 4083 /* check flags correspondence */ 4084 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT)); 4085 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT)); 4086 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT)); 4087 BUILD_BUG_ON(SKB_GSO_TCP_FIXEDID != (NETIF_F_TSO_MANGLEID >> NETIF_F_GSO_SHIFT)); 4088 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT)); 4089 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT)); 4090 BUILD_BUG_ON(SKB_GSO_GRE != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT)); 4091 BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT)); 4092 BUILD_BUG_ON(SKB_GSO_IPXIP4 != (NETIF_F_GSO_IPXIP4 >> NETIF_F_GSO_SHIFT)); 4093 BUILD_BUG_ON(SKB_GSO_IPXIP6 != (NETIF_F_GSO_IPXIP6 >> NETIF_F_GSO_SHIFT)); 4094 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT)); 4095 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT)); 4096 BUILD_BUG_ON(SKB_GSO_PARTIAL != (NETIF_F_GSO_PARTIAL >> NETIF_F_GSO_SHIFT)); 4097 BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT)); 4098 BUILD_BUG_ON(SKB_GSO_SCTP != (NETIF_F_GSO_SCTP >> NETIF_F_GSO_SHIFT)); 4099 BUILD_BUG_ON(SKB_GSO_ESP != (NETIF_F_GSO_ESP >> NETIF_F_GSO_SHIFT)); 4100 4101 return (features & feature) == feature; 4102 } 4103 4104 static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features) 4105 { 4106 return net_gso_ok(features, skb_shinfo(skb)->gso_type) && 4107 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST)); 4108 } 4109 4110 static inline bool netif_needs_gso(struct sk_buff *skb, 4111 netdev_features_t features) 4112 { 4113 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) || 4114 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) && 4115 (skb->ip_summed != CHECKSUM_UNNECESSARY))); 4116 } 4117 4118 static inline void netif_set_gso_max_size(struct net_device *dev, 4119 unsigned int size) 4120 { 4121 dev->gso_max_size = size; 4122 } 4123 4124 static inline void skb_gso_error_unwind(struct sk_buff *skb, __be16 protocol, 4125 int pulled_hlen, u16 mac_offset, 4126 int mac_len) 4127 { 4128 skb->protocol = protocol; 4129 skb->encapsulation = 1; 4130 skb_push(skb, pulled_hlen); 4131 skb_reset_transport_header(skb); 4132 skb->mac_header = mac_offset; 4133 skb->network_header = skb->mac_header + mac_len; 4134 skb->mac_len = mac_len; 4135 } 4136 4137 static inline bool netif_is_macsec(const struct net_device *dev) 4138 { 4139 return dev->priv_flags & IFF_MACSEC; 4140 } 4141 4142 static inline bool netif_is_macvlan(const struct net_device *dev) 4143 { 4144 return dev->priv_flags & IFF_MACVLAN; 4145 } 4146 4147 static inline bool netif_is_macvlan_port(const struct net_device *dev) 4148 { 4149 return dev->priv_flags & IFF_MACVLAN_PORT; 4150 } 4151 4152 static inline bool netif_is_ipvlan(const struct net_device *dev) 4153 { 4154 return dev->priv_flags & IFF_IPVLAN_SLAVE; 4155 } 4156 4157 static inline bool netif_is_ipvlan_port(const struct net_device *dev) 4158 { 4159 return dev->priv_flags & IFF_IPVLAN_MASTER; 4160 } 4161 4162 static inline bool netif_is_bond_master(const struct net_device *dev) 4163 { 4164 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING; 4165 } 4166 4167 static inline bool netif_is_bond_slave(const struct net_device *dev) 4168 { 4169 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING; 4170 } 4171 4172 static inline bool netif_supports_nofcs(struct net_device *dev) 4173 { 4174 return dev->priv_flags & IFF_SUPP_NOFCS; 4175 } 4176 4177 static inline bool netif_is_l3_master(const struct net_device *dev) 4178 { 4179 return dev->priv_flags & IFF_L3MDEV_MASTER; 4180 } 4181 4182 static inline bool netif_is_l3_slave(const struct net_device *dev) 4183 { 4184 return dev->priv_flags & IFF_L3MDEV_SLAVE; 4185 } 4186 4187 static inline bool netif_is_bridge_master(const struct net_device *dev) 4188 { 4189 return dev->priv_flags & IFF_EBRIDGE; 4190 } 4191 4192 static inline bool netif_is_bridge_port(const struct net_device *dev) 4193 { 4194 return dev->priv_flags & IFF_BRIDGE_PORT; 4195 } 4196 4197 static inline bool netif_is_ovs_master(const struct net_device *dev) 4198 { 4199 return dev->priv_flags & IFF_OPENVSWITCH; 4200 } 4201 4202 static inline bool netif_is_ovs_port(const struct net_device *dev) 4203 { 4204 return dev->priv_flags & IFF_OVS_DATAPATH; 4205 } 4206 4207 static inline bool netif_is_team_master(const struct net_device *dev) 4208 { 4209 return dev->priv_flags & IFF_TEAM; 4210 } 4211 4212 static inline bool netif_is_team_port(const struct net_device *dev) 4213 { 4214 return dev->priv_flags & IFF_TEAM_PORT; 4215 } 4216 4217 static inline bool netif_is_lag_master(const struct net_device *dev) 4218 { 4219 return netif_is_bond_master(dev) || netif_is_team_master(dev); 4220 } 4221 4222 static inline bool netif_is_lag_port(const struct net_device *dev) 4223 { 4224 return netif_is_bond_slave(dev) || netif_is_team_port(dev); 4225 } 4226 4227 static inline bool netif_is_rxfh_configured(const struct net_device *dev) 4228 { 4229 return dev->priv_flags & IFF_RXFH_CONFIGURED; 4230 } 4231 4232 /* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */ 4233 static inline void netif_keep_dst(struct net_device *dev) 4234 { 4235 dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM); 4236 } 4237 4238 /* return true if dev can't cope with mtu frames that need vlan tag insertion */ 4239 static inline bool netif_reduces_vlan_mtu(struct net_device *dev) 4240 { 4241 /* TODO: reserve and use an additional IFF bit, if we get more users */ 4242 return dev->priv_flags & IFF_MACSEC; 4243 } 4244 4245 extern struct pernet_operations __net_initdata loopback_net_ops; 4246 4247 /* Logging, debugging and troubleshooting/diagnostic helpers. */ 4248 4249 /* netdev_printk helpers, similar to dev_printk */ 4250 4251 static inline const char *netdev_name(const struct net_device *dev) 4252 { 4253 if (!dev->name[0] || strchr(dev->name, '%')) 4254 return "(unnamed net_device)"; 4255 return dev->name; 4256 } 4257 4258 static inline bool netdev_unregistering(const struct net_device *dev) 4259 { 4260 return dev->reg_state == NETREG_UNREGISTERING; 4261 } 4262 4263 static inline const char *netdev_reg_state(const struct net_device *dev) 4264 { 4265 switch (dev->reg_state) { 4266 case NETREG_UNINITIALIZED: return " (uninitialized)"; 4267 case NETREG_REGISTERED: return ""; 4268 case NETREG_UNREGISTERING: return " (unregistering)"; 4269 case NETREG_UNREGISTERED: return " (unregistered)"; 4270 case NETREG_RELEASED: return " (released)"; 4271 case NETREG_DUMMY: return " (dummy)"; 4272 } 4273 4274 WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, dev->reg_state); 4275 return " (unknown)"; 4276 } 4277 4278 __printf(3, 4) 4279 void netdev_printk(const char *level, const struct net_device *dev, 4280 const char *format, ...); 4281 __printf(2, 3) 4282 void netdev_emerg(const struct net_device *dev, const char *format, ...); 4283 __printf(2, 3) 4284 void netdev_alert(const struct net_device *dev, const char *format, ...); 4285 __printf(2, 3) 4286 void netdev_crit(const struct net_device *dev, const char *format, ...); 4287 __printf(2, 3) 4288 void netdev_err(const struct net_device *dev, const char *format, ...); 4289 __printf(2, 3) 4290 void netdev_warn(const struct net_device *dev, const char *format, ...); 4291 __printf(2, 3) 4292 void netdev_notice(const struct net_device *dev, const char *format, ...); 4293 __printf(2, 3) 4294 void netdev_info(const struct net_device *dev, const char *format, ...); 4295 4296 #define MODULE_ALIAS_NETDEV(device) \ 4297 MODULE_ALIAS("netdev-" device) 4298 4299 #if defined(CONFIG_DYNAMIC_DEBUG) 4300 #define netdev_dbg(__dev, format, args...) \ 4301 do { \ 4302 dynamic_netdev_dbg(__dev, format, ##args); \ 4303 } while (0) 4304 #elif defined(DEBUG) 4305 #define netdev_dbg(__dev, format, args...) \ 4306 netdev_printk(KERN_DEBUG, __dev, format, ##args) 4307 #else 4308 #define netdev_dbg(__dev, format, args...) \ 4309 ({ \ 4310 if (0) \ 4311 netdev_printk(KERN_DEBUG, __dev, format, ##args); \ 4312 }) 4313 #endif 4314 4315 #if defined(VERBOSE_DEBUG) 4316 #define netdev_vdbg netdev_dbg 4317 #else 4318 4319 #define netdev_vdbg(dev, format, args...) \ 4320 ({ \ 4321 if (0) \ 4322 netdev_printk(KERN_DEBUG, dev, format, ##args); \ 4323 0; \ 4324 }) 4325 #endif 4326 4327 /* 4328 * netdev_WARN() acts like dev_printk(), but with the key difference 4329 * of using a WARN/WARN_ON to get the message out, including the 4330 * file/line information and a backtrace. 4331 */ 4332 #define netdev_WARN(dev, format, args...) \ 4333 WARN(1, "netdevice: %s%s\n" format, netdev_name(dev), \ 4334 netdev_reg_state(dev), ##args) 4335 4336 /* netif printk helpers, similar to netdev_printk */ 4337 4338 #define netif_printk(priv, type, level, dev, fmt, args...) \ 4339 do { \ 4340 if (netif_msg_##type(priv)) \ 4341 netdev_printk(level, (dev), fmt, ##args); \ 4342 } while (0) 4343 4344 #define netif_level(level, priv, type, dev, fmt, args...) \ 4345 do { \ 4346 if (netif_msg_##type(priv)) \ 4347 netdev_##level(dev, fmt, ##args); \ 4348 } while (0) 4349 4350 #define netif_emerg(priv, type, dev, fmt, args...) \ 4351 netif_level(emerg, priv, type, dev, fmt, ##args) 4352 #define netif_alert(priv, type, dev, fmt, args...) \ 4353 netif_level(alert, priv, type, dev, fmt, ##args) 4354 #define netif_crit(priv, type, dev, fmt, args...) \ 4355 netif_level(crit, priv, type, dev, fmt, ##args) 4356 #define netif_err(priv, type, dev, fmt, args...) \ 4357 netif_level(err, priv, type, dev, fmt, ##args) 4358 #define netif_warn(priv, type, dev, fmt, args...) \ 4359 netif_level(warn, priv, type, dev, fmt, ##args) 4360 #define netif_notice(priv, type, dev, fmt, args...) \ 4361 netif_level(notice, priv, type, dev, fmt, ##args) 4362 #define netif_info(priv, type, dev, fmt, args...) \ 4363 netif_level(info, priv, type, dev, fmt, ##args) 4364 4365 #if defined(CONFIG_DYNAMIC_DEBUG) 4366 #define netif_dbg(priv, type, netdev, format, args...) \ 4367 do { \ 4368 if (netif_msg_##type(priv)) \ 4369 dynamic_netdev_dbg(netdev, format, ##args); \ 4370 } while (0) 4371 #elif defined(DEBUG) 4372 #define netif_dbg(priv, type, dev, format, args...) \ 4373 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args) 4374 #else 4375 #define netif_dbg(priv, type, dev, format, args...) \ 4376 ({ \ 4377 if (0) \ 4378 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \ 4379 0; \ 4380 }) 4381 #endif 4382 4383 /* if @cond then downgrade to debug, else print at @level */ 4384 #define netif_cond_dbg(priv, type, netdev, cond, level, fmt, args...) \ 4385 do { \ 4386 if (cond) \ 4387 netif_dbg(priv, type, netdev, fmt, ##args); \ 4388 else \ 4389 netif_ ## level(priv, type, netdev, fmt, ##args); \ 4390 } while (0) 4391 4392 #if defined(VERBOSE_DEBUG) 4393 #define netif_vdbg netif_dbg 4394 #else 4395 #define netif_vdbg(priv, type, dev, format, args...) \ 4396 ({ \ 4397 if (0) \ 4398 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \ 4399 0; \ 4400 }) 4401 #endif 4402 4403 /* 4404 * The list of packet types we will receive (as opposed to discard) 4405 * and the routines to invoke. 4406 * 4407 * Why 16. Because with 16 the only overlap we get on a hash of the 4408 * low nibble of the protocol value is RARP/SNAP/X.25. 4409 * 4410 * NOTE: That is no longer true with the addition of VLAN tags. Not 4411 * sure which should go first, but I bet it won't make much 4412 * difference if we are running VLANs. The good news is that 4413 * this protocol won't be in the list unless compiled in, so 4414 * the average user (w/out VLANs) will not be adversely affected. 4415 * --BLG 4416 * 4417 * 0800 IP 4418 * 8100 802.1Q VLAN 4419 * 0001 802.3 4420 * 0002 AX.25 4421 * 0004 802.2 4422 * 8035 RARP 4423 * 0005 SNAP 4424 * 0805 X.25 4425 * 0806 ARP 4426 * 8137 IPX 4427 * 0009 Localtalk 4428 * 86DD IPv6 4429 */ 4430 #define PTYPE_HASH_SIZE (16) 4431 #define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1) 4432 4433 #endif /* _LINUX_NETDEVICE_H */ 4434