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/if.h> 29 #include <linux/if_ether.h> 30 #include <linux/if_packet.h> 31 #include <linux/if_link.h> 32 33 #ifdef __KERNEL__ 34 #include <linux/pm_qos.h> 35 #include <linux/timer.h> 36 #include <linux/bug.h> 37 #include <linux/delay.h> 38 #include <linux/atomic.h> 39 #include <asm/cache.h> 40 #include <asm/byteorder.h> 41 42 #include <linux/percpu.h> 43 #include <linux/rculist.h> 44 #include <linux/dmaengine.h> 45 #include <linux/workqueue.h> 46 #include <linux/dynamic_queue_limits.h> 47 48 #include <linux/ethtool.h> 49 #include <net/net_namespace.h> 50 #include <net/dsa.h> 51 #ifdef CONFIG_DCB 52 #include <net/dcbnl.h> 53 #endif 54 #include <net/netprio_cgroup.h> 55 56 #include <linux/netdev_features.h> 57 #include <linux/neighbour.h> 58 59 struct netpoll_info; 60 struct device; 61 struct phy_device; 62 /* 802.11 specific */ 63 struct wireless_dev; 64 /* source back-compat hooks */ 65 #define SET_ETHTOOL_OPS(netdev,ops) \ 66 ( (netdev)->ethtool_ops = (ops) ) 67 68 /* hardware address assignment types */ 69 #define NET_ADDR_PERM 0 /* address is permanent (default) */ 70 #define NET_ADDR_RANDOM 1 /* address is generated randomly */ 71 #define NET_ADDR_STOLEN 2 /* address is stolen from other device */ 72 73 /* Backlog congestion levels */ 74 #define NET_RX_SUCCESS 0 /* keep 'em coming, baby */ 75 #define NET_RX_DROP 1 /* packet dropped */ 76 77 /* 78 * Transmit return codes: transmit return codes originate from three different 79 * namespaces: 80 * 81 * - qdisc return codes 82 * - driver transmit return codes 83 * - errno values 84 * 85 * Drivers are allowed to return any one of those in their hard_start_xmit() 86 * function. Real network devices commonly used with qdiscs should only return 87 * the driver transmit return codes though - when qdiscs are used, the actual 88 * transmission happens asynchronously, so the value is not propagated to 89 * higher layers. Virtual network devices transmit synchronously, in this case 90 * the driver transmit return codes are consumed by dev_queue_xmit(), all 91 * others are propagated to higher layers. 92 */ 93 94 /* qdisc ->enqueue() return codes. */ 95 #define NET_XMIT_SUCCESS 0x00 96 #define NET_XMIT_DROP 0x01 /* skb dropped */ 97 #define NET_XMIT_CN 0x02 /* congestion notification */ 98 #define NET_XMIT_POLICED 0x03 /* skb is shot by police */ 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 NETDEV_TX_LOCKED = 0x20, /* driver tx lock was already taken */ 115 }; 116 typedef enum netdev_tx netdev_tx_t; 117 118 /* 119 * Current order: NETDEV_TX_MASK > NET_XMIT_MASK >= 0 is significant; 120 * hard_start_xmit() return < NET_XMIT_MASK means skb was consumed. 121 */ 122 static inline bool dev_xmit_complete(int rc) 123 { 124 /* 125 * Positive cases with an skb consumed by a driver: 126 * - successful transmission (rc == NETDEV_TX_OK) 127 * - error while transmitting (rc < 0) 128 * - error while queueing to a different device (rc & NET_XMIT_MASK) 129 */ 130 if (likely(rc < NET_XMIT_MASK)) 131 return true; 132 133 return false; 134 } 135 136 #endif 137 138 #define MAX_ADDR_LEN 32 /* Largest hardware address length */ 139 140 /* Initial net device group. All devices belong to group 0 by default. */ 141 #define INIT_NETDEV_GROUP 0 142 143 #ifdef __KERNEL__ 144 /* 145 * Compute the worst case header length according to the protocols 146 * used. 147 */ 148 149 #if defined(CONFIG_WLAN) || IS_ENABLED(CONFIG_AX25) 150 # if defined(CONFIG_MAC80211_MESH) 151 # define LL_MAX_HEADER 128 152 # else 153 # define LL_MAX_HEADER 96 154 # endif 155 #elif IS_ENABLED(CONFIG_TR) 156 # define LL_MAX_HEADER 48 157 #else 158 # define LL_MAX_HEADER 32 159 #endif 160 161 #if !IS_ENABLED(CONFIG_NET_IPIP) && !IS_ENABLED(CONFIG_NET_IPGRE) && \ 162 !IS_ENABLED(CONFIG_IPV6_SIT) && !IS_ENABLED(CONFIG_IPV6_TUNNEL) 163 #define MAX_HEADER LL_MAX_HEADER 164 #else 165 #define MAX_HEADER (LL_MAX_HEADER + 48) 166 #endif 167 168 /* 169 * Old network device statistics. Fields are native words 170 * (unsigned long) so they can be read and written atomically. 171 */ 172 173 struct net_device_stats { 174 unsigned long rx_packets; 175 unsigned long tx_packets; 176 unsigned long rx_bytes; 177 unsigned long tx_bytes; 178 unsigned long rx_errors; 179 unsigned long tx_errors; 180 unsigned long rx_dropped; 181 unsigned long tx_dropped; 182 unsigned long multicast; 183 unsigned long collisions; 184 unsigned long rx_length_errors; 185 unsigned long rx_over_errors; 186 unsigned long rx_crc_errors; 187 unsigned long rx_frame_errors; 188 unsigned long rx_fifo_errors; 189 unsigned long rx_missed_errors; 190 unsigned long tx_aborted_errors; 191 unsigned long tx_carrier_errors; 192 unsigned long tx_fifo_errors; 193 unsigned long tx_heartbeat_errors; 194 unsigned long tx_window_errors; 195 unsigned long rx_compressed; 196 unsigned long tx_compressed; 197 }; 198 199 #endif /* __KERNEL__ */ 200 201 202 /* Media selection options. */ 203 enum { 204 IF_PORT_UNKNOWN = 0, 205 IF_PORT_10BASE2, 206 IF_PORT_10BASET, 207 IF_PORT_AUI, 208 IF_PORT_100BASET, 209 IF_PORT_100BASETX, 210 IF_PORT_100BASEFX 211 }; 212 213 #ifdef __KERNEL__ 214 215 #include <linux/cache.h> 216 #include <linux/skbuff.h> 217 218 #ifdef CONFIG_RPS 219 #include <linux/static_key.h> 220 extern struct static_key rps_needed; 221 #endif 222 223 struct neighbour; 224 struct neigh_parms; 225 struct sk_buff; 226 227 struct netdev_hw_addr { 228 struct list_head list; 229 unsigned char addr[MAX_ADDR_LEN]; 230 unsigned char type; 231 #define NETDEV_HW_ADDR_T_LAN 1 232 #define NETDEV_HW_ADDR_T_SAN 2 233 #define NETDEV_HW_ADDR_T_SLAVE 3 234 #define NETDEV_HW_ADDR_T_UNICAST 4 235 #define NETDEV_HW_ADDR_T_MULTICAST 5 236 bool synced; 237 bool global_use; 238 int refcount; 239 struct rcu_head rcu_head; 240 }; 241 242 struct netdev_hw_addr_list { 243 struct list_head list; 244 int count; 245 }; 246 247 #define netdev_hw_addr_list_count(l) ((l)->count) 248 #define netdev_hw_addr_list_empty(l) (netdev_hw_addr_list_count(l) == 0) 249 #define netdev_hw_addr_list_for_each(ha, l) \ 250 list_for_each_entry(ha, &(l)->list, list) 251 252 #define netdev_uc_count(dev) netdev_hw_addr_list_count(&(dev)->uc) 253 #define netdev_uc_empty(dev) netdev_hw_addr_list_empty(&(dev)->uc) 254 #define netdev_for_each_uc_addr(ha, dev) \ 255 netdev_hw_addr_list_for_each(ha, &(dev)->uc) 256 257 #define netdev_mc_count(dev) netdev_hw_addr_list_count(&(dev)->mc) 258 #define netdev_mc_empty(dev) netdev_hw_addr_list_empty(&(dev)->mc) 259 #define netdev_for_each_mc_addr(ha, dev) \ 260 netdev_hw_addr_list_for_each(ha, &(dev)->mc) 261 262 struct hh_cache { 263 u16 hh_len; 264 u16 __pad; 265 seqlock_t hh_lock; 266 267 /* cached hardware header; allow for machine alignment needs. */ 268 #define HH_DATA_MOD 16 269 #define HH_DATA_OFF(__len) \ 270 (HH_DATA_MOD - (((__len - 1) & (HH_DATA_MOD - 1)) + 1)) 271 #define HH_DATA_ALIGN(__len) \ 272 (((__len)+(HH_DATA_MOD-1))&~(HH_DATA_MOD - 1)) 273 unsigned long hh_data[HH_DATA_ALIGN(LL_MAX_HEADER) / sizeof(long)]; 274 }; 275 276 /* Reserve HH_DATA_MOD byte aligned hard_header_len, but at least that much. 277 * Alternative is: 278 * dev->hard_header_len ? (dev->hard_header_len + 279 * (HH_DATA_MOD - 1)) & ~(HH_DATA_MOD - 1) : 0 280 * 281 * We could use other alignment values, but we must maintain the 282 * relationship HH alignment <= LL alignment. 283 */ 284 #define LL_RESERVED_SPACE(dev) \ 285 ((((dev)->hard_header_len+(dev)->needed_headroom)&~(HH_DATA_MOD - 1)) + HH_DATA_MOD) 286 #define LL_RESERVED_SPACE_EXTRA(dev,extra) \ 287 ((((dev)->hard_header_len+(dev)->needed_headroom+(extra))&~(HH_DATA_MOD - 1)) + HH_DATA_MOD) 288 289 struct header_ops { 290 int (*create) (struct sk_buff *skb, struct net_device *dev, 291 unsigned short type, const void *daddr, 292 const void *saddr, unsigned int len); 293 int (*parse)(const struct sk_buff *skb, unsigned char *haddr); 294 int (*rebuild)(struct sk_buff *skb); 295 int (*cache)(const struct neighbour *neigh, struct hh_cache *hh, __be16 type); 296 void (*cache_update)(struct hh_cache *hh, 297 const struct net_device *dev, 298 const unsigned char *haddr); 299 }; 300 301 /* These flag bits are private to the generic network queueing 302 * layer, they may not be explicitly referenced by any other 303 * code. 304 */ 305 306 enum netdev_state_t { 307 __LINK_STATE_START, 308 __LINK_STATE_PRESENT, 309 __LINK_STATE_NOCARRIER, 310 __LINK_STATE_LINKWATCH_PENDING, 311 __LINK_STATE_DORMANT, 312 }; 313 314 315 /* 316 * This structure holds at boot time configured netdevice settings. They 317 * are then used in the device probing. 318 */ 319 struct netdev_boot_setup { 320 char name[IFNAMSIZ]; 321 struct ifmap map; 322 }; 323 #define NETDEV_BOOT_SETUP_MAX 8 324 325 extern int __init netdev_boot_setup(char *str); 326 327 /* 328 * Structure for NAPI scheduling similar to tasklet but with weighting 329 */ 330 struct napi_struct { 331 /* The poll_list must only be managed by the entity which 332 * changes the state of the NAPI_STATE_SCHED bit. This means 333 * whoever atomically sets that bit can add this napi_struct 334 * to the per-cpu poll_list, and whoever clears that bit 335 * can remove from the list right before clearing the bit. 336 */ 337 struct list_head poll_list; 338 339 unsigned long state; 340 int weight; 341 unsigned int gro_count; 342 int (*poll)(struct napi_struct *, int); 343 #ifdef CONFIG_NETPOLL 344 spinlock_t poll_lock; 345 int poll_owner; 346 #endif 347 struct net_device *dev; 348 struct sk_buff *gro_list; 349 struct sk_buff *skb; 350 struct list_head dev_list; 351 }; 352 353 enum { 354 NAPI_STATE_SCHED, /* Poll is scheduled */ 355 NAPI_STATE_DISABLE, /* Disable pending */ 356 NAPI_STATE_NPSVC, /* Netpoll - don't dequeue from poll_list */ 357 }; 358 359 enum gro_result { 360 GRO_MERGED, 361 GRO_MERGED_FREE, 362 GRO_HELD, 363 GRO_NORMAL, 364 GRO_DROP, 365 }; 366 typedef enum gro_result gro_result_t; 367 368 /* 369 * enum rx_handler_result - Possible return values for rx_handlers. 370 * @RX_HANDLER_CONSUMED: skb was consumed by rx_handler, do not process it 371 * further. 372 * @RX_HANDLER_ANOTHER: Do another round in receive path. This is indicated in 373 * case skb->dev was changed by rx_handler. 374 * @RX_HANDLER_EXACT: Force exact delivery, no wildcard. 375 * @RX_HANDLER_PASS: Do nothing, passe the skb as if no rx_handler was called. 376 * 377 * rx_handlers are functions called from inside __netif_receive_skb(), to do 378 * special processing of the skb, prior to delivery to protocol handlers. 379 * 380 * Currently, a net_device can only have a single rx_handler registered. Trying 381 * to register a second rx_handler will return -EBUSY. 382 * 383 * To register a rx_handler on a net_device, use netdev_rx_handler_register(). 384 * To unregister a rx_handler on a net_device, use 385 * netdev_rx_handler_unregister(). 386 * 387 * Upon return, rx_handler is expected to tell __netif_receive_skb() what to 388 * do with the skb. 389 * 390 * If the rx_handler consumed to skb in some way, it should return 391 * RX_HANDLER_CONSUMED. This is appropriate when the rx_handler arranged for 392 * the skb to be delivered in some other ways. 393 * 394 * If the rx_handler changed skb->dev, to divert the skb to another 395 * net_device, it should return RX_HANDLER_ANOTHER. The rx_handler for the 396 * new device will be called if it exists. 397 * 398 * If the rx_handler consider the skb should be ignored, it should return 399 * RX_HANDLER_EXACT. The skb will only be delivered to protocol handlers that 400 * are registred on exact device (ptype->dev == skb->dev). 401 * 402 * If the rx_handler didn't changed skb->dev, but want the skb to be normally 403 * delivered, it should return RX_HANDLER_PASS. 404 * 405 * A device without a registered rx_handler will behave as if rx_handler 406 * returned RX_HANDLER_PASS. 407 */ 408 409 enum rx_handler_result { 410 RX_HANDLER_CONSUMED, 411 RX_HANDLER_ANOTHER, 412 RX_HANDLER_EXACT, 413 RX_HANDLER_PASS, 414 }; 415 typedef enum rx_handler_result rx_handler_result_t; 416 typedef rx_handler_result_t rx_handler_func_t(struct sk_buff **pskb); 417 418 extern void __napi_schedule(struct napi_struct *n); 419 420 static inline bool napi_disable_pending(struct napi_struct *n) 421 { 422 return test_bit(NAPI_STATE_DISABLE, &n->state); 423 } 424 425 /** 426 * napi_schedule_prep - check if napi can be scheduled 427 * @n: napi context 428 * 429 * Test if NAPI routine is already running, and if not mark 430 * it as running. This is used as a condition variable 431 * insure only one NAPI poll instance runs. We also make 432 * sure there is no pending NAPI disable. 433 */ 434 static inline bool napi_schedule_prep(struct napi_struct *n) 435 { 436 return !napi_disable_pending(n) && 437 !test_and_set_bit(NAPI_STATE_SCHED, &n->state); 438 } 439 440 /** 441 * napi_schedule - schedule NAPI poll 442 * @n: napi context 443 * 444 * Schedule NAPI poll routine to be called if it is not already 445 * running. 446 */ 447 static inline void napi_schedule(struct napi_struct *n) 448 { 449 if (napi_schedule_prep(n)) 450 __napi_schedule(n); 451 } 452 453 /* Try to reschedule poll. Called by dev->poll() after napi_complete(). */ 454 static inline bool napi_reschedule(struct napi_struct *napi) 455 { 456 if (napi_schedule_prep(napi)) { 457 __napi_schedule(napi); 458 return true; 459 } 460 return false; 461 } 462 463 /** 464 * napi_complete - NAPI processing complete 465 * @n: napi context 466 * 467 * Mark NAPI processing as complete. 468 */ 469 extern void __napi_complete(struct napi_struct *n); 470 extern void napi_complete(struct napi_struct *n); 471 472 /** 473 * napi_disable - prevent NAPI from scheduling 474 * @n: napi context 475 * 476 * Stop NAPI from being scheduled on this context. 477 * Waits till any outstanding processing completes. 478 */ 479 static inline void napi_disable(struct napi_struct *n) 480 { 481 set_bit(NAPI_STATE_DISABLE, &n->state); 482 while (test_and_set_bit(NAPI_STATE_SCHED, &n->state)) 483 msleep(1); 484 clear_bit(NAPI_STATE_DISABLE, &n->state); 485 } 486 487 /** 488 * napi_enable - enable NAPI scheduling 489 * @n: napi context 490 * 491 * Resume NAPI from being scheduled on this context. 492 * Must be paired with napi_disable. 493 */ 494 static inline void napi_enable(struct napi_struct *n) 495 { 496 BUG_ON(!test_bit(NAPI_STATE_SCHED, &n->state)); 497 smp_mb__before_clear_bit(); 498 clear_bit(NAPI_STATE_SCHED, &n->state); 499 } 500 501 #ifdef CONFIG_SMP 502 /** 503 * napi_synchronize - wait until NAPI is not running 504 * @n: napi context 505 * 506 * Wait until NAPI is done being scheduled on this context. 507 * Waits till any outstanding processing completes but 508 * does not disable future activations. 509 */ 510 static inline void napi_synchronize(const struct napi_struct *n) 511 { 512 while (test_bit(NAPI_STATE_SCHED, &n->state)) 513 msleep(1); 514 } 515 #else 516 # define napi_synchronize(n) barrier() 517 #endif 518 519 enum netdev_queue_state_t { 520 __QUEUE_STATE_DRV_XOFF, 521 __QUEUE_STATE_STACK_XOFF, 522 __QUEUE_STATE_FROZEN, 523 #define QUEUE_STATE_ANY_XOFF ((1 << __QUEUE_STATE_DRV_XOFF) | \ 524 (1 << __QUEUE_STATE_STACK_XOFF)) 525 #define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \ 526 (1 << __QUEUE_STATE_FROZEN)) 527 }; 528 /* 529 * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue. The 530 * netif_tx_* functions below are used to manipulate this flag. The 531 * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit 532 * queue independently. The netif_xmit_*stopped functions below are called 533 * to check if the queue has been stopped by the driver or stack (either 534 * of the XOFF bits are set in the state). Drivers should not need to call 535 * netif_xmit*stopped functions, they should only be using netif_tx_*. 536 */ 537 538 struct netdev_queue { 539 /* 540 * read mostly part 541 */ 542 struct net_device *dev; 543 struct Qdisc *qdisc; 544 struct Qdisc *qdisc_sleeping; 545 #ifdef CONFIG_SYSFS 546 struct kobject kobj; 547 #endif 548 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA) 549 int numa_node; 550 #endif 551 /* 552 * write mostly part 553 */ 554 spinlock_t _xmit_lock ____cacheline_aligned_in_smp; 555 int xmit_lock_owner; 556 /* 557 * please use this field instead of dev->trans_start 558 */ 559 unsigned long trans_start; 560 561 /* 562 * Number of TX timeouts for this queue 563 * (/sys/class/net/DEV/Q/trans_timeout) 564 */ 565 unsigned long trans_timeout; 566 567 unsigned long state; 568 569 #ifdef CONFIG_BQL 570 struct dql dql; 571 #endif 572 } ____cacheline_aligned_in_smp; 573 574 static inline int netdev_queue_numa_node_read(const struct netdev_queue *q) 575 { 576 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA) 577 return q->numa_node; 578 #else 579 return NUMA_NO_NODE; 580 #endif 581 } 582 583 static inline void netdev_queue_numa_node_write(struct netdev_queue *q, int node) 584 { 585 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA) 586 q->numa_node = node; 587 #endif 588 } 589 590 #ifdef CONFIG_RPS 591 /* 592 * This structure holds an RPS map which can be of variable length. The 593 * map is an array of CPUs. 594 */ 595 struct rps_map { 596 unsigned int len; 597 struct rcu_head rcu; 598 u16 cpus[0]; 599 }; 600 #define RPS_MAP_SIZE(_num) (sizeof(struct rps_map) + ((_num) * sizeof(u16))) 601 602 /* 603 * The rps_dev_flow structure contains the mapping of a flow to a CPU, the 604 * tail pointer for that CPU's input queue at the time of last enqueue, and 605 * a hardware filter index. 606 */ 607 struct rps_dev_flow { 608 u16 cpu; 609 u16 filter; 610 unsigned int last_qtail; 611 }; 612 #define RPS_NO_FILTER 0xffff 613 614 /* 615 * The rps_dev_flow_table structure contains a table of flow mappings. 616 */ 617 struct rps_dev_flow_table { 618 unsigned int mask; 619 struct rcu_head rcu; 620 struct work_struct free_work; 621 struct rps_dev_flow flows[0]; 622 }; 623 #define RPS_DEV_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_dev_flow_table) + \ 624 ((_num) * sizeof(struct rps_dev_flow))) 625 626 /* 627 * The rps_sock_flow_table contains mappings of flows to the last CPU 628 * on which they were processed by the application (set in recvmsg). 629 */ 630 struct rps_sock_flow_table { 631 unsigned int mask; 632 u16 ents[0]; 633 }; 634 #define RPS_SOCK_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_sock_flow_table) + \ 635 ((_num) * sizeof(u16))) 636 637 #define RPS_NO_CPU 0xffff 638 639 static inline void rps_record_sock_flow(struct rps_sock_flow_table *table, 640 u32 hash) 641 { 642 if (table && hash) { 643 unsigned int cpu, index = hash & table->mask; 644 645 /* We only give a hint, preemption can change cpu under us */ 646 cpu = raw_smp_processor_id(); 647 648 if (table->ents[index] != cpu) 649 table->ents[index] = cpu; 650 } 651 } 652 653 static inline void rps_reset_sock_flow(struct rps_sock_flow_table *table, 654 u32 hash) 655 { 656 if (table && hash) 657 table->ents[hash & table->mask] = RPS_NO_CPU; 658 } 659 660 extern struct rps_sock_flow_table __rcu *rps_sock_flow_table; 661 662 #ifdef CONFIG_RFS_ACCEL 663 extern bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index, 664 u32 flow_id, u16 filter_id); 665 #endif 666 667 /* This structure contains an instance of an RX queue. */ 668 struct netdev_rx_queue { 669 struct rps_map __rcu *rps_map; 670 struct rps_dev_flow_table __rcu *rps_flow_table; 671 struct kobject kobj; 672 struct net_device *dev; 673 } ____cacheline_aligned_in_smp; 674 #endif /* CONFIG_RPS */ 675 676 #ifdef CONFIG_XPS 677 /* 678 * This structure holds an XPS map which can be of variable length. The 679 * map is an array of queues. 680 */ 681 struct xps_map { 682 unsigned int len; 683 unsigned int alloc_len; 684 struct rcu_head rcu; 685 u16 queues[0]; 686 }; 687 #define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16))) 688 #define XPS_MIN_MAP_ALLOC ((L1_CACHE_BYTES - sizeof(struct xps_map)) \ 689 / sizeof(u16)) 690 691 /* 692 * This structure holds all XPS maps for device. Maps are indexed by CPU. 693 */ 694 struct xps_dev_maps { 695 struct rcu_head rcu; 696 struct xps_map __rcu *cpu_map[0]; 697 }; 698 #define XPS_DEV_MAPS_SIZE (sizeof(struct xps_dev_maps) + \ 699 (nr_cpu_ids * sizeof(struct xps_map *))) 700 #endif /* CONFIG_XPS */ 701 702 #define TC_MAX_QUEUE 16 703 #define TC_BITMASK 15 704 /* HW offloaded queuing disciplines txq count and offset maps */ 705 struct netdev_tc_txq { 706 u16 count; 707 u16 offset; 708 }; 709 710 #if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE) 711 /* 712 * This structure is to hold information about the device 713 * configured to run FCoE protocol stack. 714 */ 715 struct netdev_fcoe_hbainfo { 716 char manufacturer[64]; 717 char serial_number[64]; 718 char hardware_version[64]; 719 char driver_version[64]; 720 char optionrom_version[64]; 721 char firmware_version[64]; 722 char model[256]; 723 char model_description[256]; 724 }; 725 #endif 726 727 /* 728 * This structure defines the management hooks for network devices. 729 * The following hooks can be defined; unless noted otherwise, they are 730 * optional and can be filled with a null pointer. 731 * 732 * int (*ndo_init)(struct net_device *dev); 733 * This function is called once when network device is registered. 734 * The network device can use this to any late stage initializaton 735 * or semantic validattion. It can fail with an error code which will 736 * be propogated back to register_netdev 737 * 738 * void (*ndo_uninit)(struct net_device *dev); 739 * This function is called when device is unregistered or when registration 740 * fails. It is not called if init fails. 741 * 742 * int (*ndo_open)(struct net_device *dev); 743 * This function is called when network device transistions to the up 744 * state. 745 * 746 * int (*ndo_stop)(struct net_device *dev); 747 * This function is called when network device transistions to the down 748 * state. 749 * 750 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb, 751 * struct net_device *dev); 752 * Called when a packet needs to be transmitted. 753 * Must return NETDEV_TX_OK , NETDEV_TX_BUSY. 754 * (can also return NETDEV_TX_LOCKED iff NETIF_F_LLTX) 755 * Required can not be NULL. 756 * 757 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb); 758 * Called to decide which queue to when device supports multiple 759 * transmit queues. 760 * 761 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags); 762 * This function is called to allow device receiver to make 763 * changes to configuration when multicast or promiscious is enabled. 764 * 765 * void (*ndo_set_rx_mode)(struct net_device *dev); 766 * This function is called device changes address list filtering. 767 * If driver handles unicast address filtering, it should set 768 * IFF_UNICAST_FLT to its priv_flags. 769 * 770 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr); 771 * This function is called when the Media Access Control address 772 * needs to be changed. If this interface is not defined, the 773 * mac address can not be changed. 774 * 775 * int (*ndo_validate_addr)(struct net_device *dev); 776 * Test if Media Access Control address is valid for the device. 777 * 778 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd); 779 * Called when a user request an ioctl which can't be handled by 780 * the generic interface code. If not defined ioctl's return 781 * not supported error code. 782 * 783 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map); 784 * Used to set network devices bus interface parameters. This interface 785 * is retained for legacy reason, new devices should use the bus 786 * interface (PCI) for low level management. 787 * 788 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu); 789 * Called when a user wants to change the Maximum Transfer Unit 790 * of a device. If not defined, any request to change MTU will 791 * will return an error. 792 * 793 * void (*ndo_tx_timeout)(struct net_device *dev); 794 * Callback uses when the transmitter has not made any progress 795 * for dev->watchdog ticks. 796 * 797 * struct rtnl_link_stats64* (*ndo_get_stats64)(struct net_device *dev, 798 * struct rtnl_link_stats64 *storage); 799 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev); 800 * Called when a user wants to get the network device usage 801 * statistics. Drivers must do one of the following: 802 * 1. Define @ndo_get_stats64 to fill in a zero-initialised 803 * rtnl_link_stats64 structure passed by the caller. 804 * 2. Define @ndo_get_stats to update a net_device_stats structure 805 * (which should normally be dev->stats) and return a pointer to 806 * it. The structure may be changed asynchronously only if each 807 * field is written atomically. 808 * 3. Update dev->stats asynchronously and atomically, and define 809 * neither operation. 810 * 811 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, unsigned short vid); 812 * If device support VLAN filtering (dev->features & NETIF_F_HW_VLAN_FILTER) 813 * this function is called when a VLAN id is registered. 814 * 815 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, unsigned short vid); 816 * If device support VLAN filtering (dev->features & NETIF_F_HW_VLAN_FILTER) 817 * this function is called when a VLAN id is unregistered. 818 * 819 * void (*ndo_poll_controller)(struct net_device *dev); 820 * 821 * SR-IOV management functions. 822 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac); 823 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan, u8 qos); 824 * int (*ndo_set_vf_tx_rate)(struct net_device *dev, int vf, int rate); 825 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting); 826 * int (*ndo_get_vf_config)(struct net_device *dev, 827 * int vf, struct ifla_vf_info *ivf); 828 * int (*ndo_set_vf_port)(struct net_device *dev, int vf, 829 * struct nlattr *port[]); 830 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb); 831 * int (*ndo_setup_tc)(struct net_device *dev, u8 tc) 832 * Called to setup 'tc' number of traffic classes in the net device. This 833 * is always called from the stack with the rtnl lock held and netif tx 834 * queues stopped. This allows the netdevice to perform queue management 835 * safely. 836 * 837 * Fiber Channel over Ethernet (FCoE) offload functions. 838 * int (*ndo_fcoe_enable)(struct net_device *dev); 839 * Called when the FCoE protocol stack wants to start using LLD for FCoE 840 * so the underlying device can perform whatever needed configuration or 841 * initialization to support acceleration of FCoE traffic. 842 * 843 * int (*ndo_fcoe_disable)(struct net_device *dev); 844 * Called when the FCoE protocol stack wants to stop using LLD for FCoE 845 * so the underlying device can perform whatever needed clean-ups to 846 * stop supporting acceleration of FCoE traffic. 847 * 848 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid, 849 * struct scatterlist *sgl, unsigned int sgc); 850 * Called when the FCoE Initiator wants to initialize an I/O that 851 * is a possible candidate for Direct Data Placement (DDP). The LLD can 852 * perform necessary setup and returns 1 to indicate the device is set up 853 * successfully to perform DDP on this I/O, otherwise this returns 0. 854 * 855 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid); 856 * Called when the FCoE Initiator/Target is done with the DDPed I/O as 857 * indicated by the FC exchange id 'xid', so the underlying device can 858 * clean up and reuse resources for later DDP requests. 859 * 860 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid, 861 * struct scatterlist *sgl, unsigned int sgc); 862 * Called when the FCoE Target wants to initialize an I/O that 863 * is a possible candidate for Direct Data Placement (DDP). The LLD can 864 * perform necessary setup and returns 1 to indicate the device is set up 865 * successfully to perform DDP on this I/O, otherwise this returns 0. 866 * 867 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev, 868 * struct netdev_fcoe_hbainfo *hbainfo); 869 * Called when the FCoE Protocol stack wants information on the underlying 870 * device. This information is utilized by the FCoE protocol stack to 871 * register attributes with Fiber Channel management service as per the 872 * FC-GS Fabric Device Management Information(FDMI) specification. 873 * 874 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type); 875 * Called when the underlying device wants to override default World Wide 876 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own 877 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE 878 * protocol stack to use. 879 * 880 * RFS acceleration. 881 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb, 882 * u16 rxq_index, u32 flow_id); 883 * Set hardware filter for RFS. rxq_index is the target queue index; 884 * flow_id is a flow ID to be passed to rps_may_expire_flow() later. 885 * Return the filter ID on success, or a negative error code. 886 * 887 * Slave management functions (for bridge, bonding, etc). User should 888 * call netdev_set_master() to set dev->master properly. 889 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev); 890 * Called to make another netdev an underling. 891 * 892 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev); 893 * Called to release previously enslaved netdev. 894 * 895 * Feature/offload setting functions. 896 * netdev_features_t (*ndo_fix_features)(struct net_device *dev, 897 * netdev_features_t features); 898 * Adjusts the requested feature flags according to device-specific 899 * constraints, and returns the resulting flags. Must not modify 900 * the device state. 901 * 902 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features); 903 * Called to update device configuration to new features. Passed 904 * feature set might be less than what was returned by ndo_fix_features()). 905 * Must return >0 or -errno if it changed dev->features itself. 906 * 907 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[], 908 * struct net_device *dev, 909 * const unsigned char *addr, u16 flags) 910 * Adds an FDB entry to dev for addr. 911 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct net_device *dev, 912 * const unsigned char *addr) 913 * Deletes the FDB entry from dev coresponding to addr. 914 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb, 915 * struct net_device *dev, int idx) 916 * Used to add FDB entries to dump requests. Implementers should add 917 * entries to skb and update idx with the number of entries. 918 */ 919 struct net_device_ops { 920 int (*ndo_init)(struct net_device *dev); 921 void (*ndo_uninit)(struct net_device *dev); 922 int (*ndo_open)(struct net_device *dev); 923 int (*ndo_stop)(struct net_device *dev); 924 netdev_tx_t (*ndo_start_xmit) (struct sk_buff *skb, 925 struct net_device *dev); 926 u16 (*ndo_select_queue)(struct net_device *dev, 927 struct sk_buff *skb); 928 void (*ndo_change_rx_flags)(struct net_device *dev, 929 int flags); 930 void (*ndo_set_rx_mode)(struct net_device *dev); 931 int (*ndo_set_mac_address)(struct net_device *dev, 932 void *addr); 933 int (*ndo_validate_addr)(struct net_device *dev); 934 int (*ndo_do_ioctl)(struct net_device *dev, 935 struct ifreq *ifr, int cmd); 936 int (*ndo_set_config)(struct net_device *dev, 937 struct ifmap *map); 938 int (*ndo_change_mtu)(struct net_device *dev, 939 int new_mtu); 940 int (*ndo_neigh_setup)(struct net_device *dev, 941 struct neigh_parms *); 942 void (*ndo_tx_timeout) (struct net_device *dev); 943 944 struct rtnl_link_stats64* (*ndo_get_stats64)(struct net_device *dev, 945 struct rtnl_link_stats64 *storage); 946 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev); 947 948 int (*ndo_vlan_rx_add_vid)(struct net_device *dev, 949 unsigned short vid); 950 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, 951 unsigned short vid); 952 #ifdef CONFIG_NET_POLL_CONTROLLER 953 void (*ndo_poll_controller)(struct net_device *dev); 954 int (*ndo_netpoll_setup)(struct net_device *dev, 955 struct netpoll_info *info, 956 gfp_t gfp); 957 void (*ndo_netpoll_cleanup)(struct net_device *dev); 958 #endif 959 int (*ndo_set_vf_mac)(struct net_device *dev, 960 int queue, u8 *mac); 961 int (*ndo_set_vf_vlan)(struct net_device *dev, 962 int queue, u16 vlan, u8 qos); 963 int (*ndo_set_vf_tx_rate)(struct net_device *dev, 964 int vf, int rate); 965 int (*ndo_set_vf_spoofchk)(struct net_device *dev, 966 int vf, bool setting); 967 int (*ndo_get_vf_config)(struct net_device *dev, 968 int vf, 969 struct ifla_vf_info *ivf); 970 int (*ndo_set_vf_port)(struct net_device *dev, 971 int vf, 972 struct nlattr *port[]); 973 int (*ndo_get_vf_port)(struct net_device *dev, 974 int vf, struct sk_buff *skb); 975 int (*ndo_setup_tc)(struct net_device *dev, u8 tc); 976 #if IS_ENABLED(CONFIG_FCOE) 977 int (*ndo_fcoe_enable)(struct net_device *dev); 978 int (*ndo_fcoe_disable)(struct net_device *dev); 979 int (*ndo_fcoe_ddp_setup)(struct net_device *dev, 980 u16 xid, 981 struct scatterlist *sgl, 982 unsigned int sgc); 983 int (*ndo_fcoe_ddp_done)(struct net_device *dev, 984 u16 xid); 985 int (*ndo_fcoe_ddp_target)(struct net_device *dev, 986 u16 xid, 987 struct scatterlist *sgl, 988 unsigned int sgc); 989 int (*ndo_fcoe_get_hbainfo)(struct net_device *dev, 990 struct netdev_fcoe_hbainfo *hbainfo); 991 #endif 992 993 #if IS_ENABLED(CONFIG_LIBFCOE) 994 #define NETDEV_FCOE_WWNN 0 995 #define NETDEV_FCOE_WWPN 1 996 int (*ndo_fcoe_get_wwn)(struct net_device *dev, 997 u64 *wwn, int type); 998 #endif 999 1000 #ifdef CONFIG_RFS_ACCEL 1001 int (*ndo_rx_flow_steer)(struct net_device *dev, 1002 const struct sk_buff *skb, 1003 u16 rxq_index, 1004 u32 flow_id); 1005 #endif 1006 int (*ndo_add_slave)(struct net_device *dev, 1007 struct net_device *slave_dev); 1008 int (*ndo_del_slave)(struct net_device *dev, 1009 struct net_device *slave_dev); 1010 netdev_features_t (*ndo_fix_features)(struct net_device *dev, 1011 netdev_features_t features); 1012 int (*ndo_set_features)(struct net_device *dev, 1013 netdev_features_t features); 1014 int (*ndo_neigh_construct)(struct neighbour *n); 1015 void (*ndo_neigh_destroy)(struct neighbour *n); 1016 1017 int (*ndo_fdb_add)(struct ndmsg *ndm, 1018 struct nlattr *tb[], 1019 struct net_device *dev, 1020 const unsigned char *addr, 1021 u16 flags); 1022 int (*ndo_fdb_del)(struct ndmsg *ndm, 1023 struct net_device *dev, 1024 const unsigned char *addr); 1025 int (*ndo_fdb_dump)(struct sk_buff *skb, 1026 struct netlink_callback *cb, 1027 struct net_device *dev, 1028 int idx); 1029 }; 1030 1031 /* 1032 * The DEVICE structure. 1033 * Actually, this whole structure is a big mistake. It mixes I/O 1034 * data with strictly "high-level" data, and it has to know about 1035 * almost every data structure used in the INET module. 1036 * 1037 * FIXME: cleanup struct net_device such that network protocol info 1038 * moves out. 1039 */ 1040 1041 struct net_device { 1042 1043 /* 1044 * This is the first field of the "visible" part of this structure 1045 * (i.e. as seen by users in the "Space.c" file). It is the name 1046 * of the interface. 1047 */ 1048 char name[IFNAMSIZ]; 1049 1050 /* device name hash chain, please keep it close to name[] */ 1051 struct hlist_node name_hlist; 1052 1053 /* snmp alias */ 1054 char *ifalias; 1055 1056 /* 1057 * I/O specific fields 1058 * FIXME: Merge these and struct ifmap into one 1059 */ 1060 unsigned long mem_end; /* shared mem end */ 1061 unsigned long mem_start; /* shared mem start */ 1062 unsigned long base_addr; /* device I/O address */ 1063 unsigned int irq; /* device IRQ number */ 1064 1065 /* 1066 * Some hardware also needs these fields, but they are not 1067 * part of the usual set specified in Space.c. 1068 */ 1069 1070 unsigned long state; 1071 1072 struct list_head dev_list; 1073 struct list_head napi_list; 1074 struct list_head unreg_list; 1075 1076 /* currently active device features */ 1077 netdev_features_t features; 1078 /* user-changeable features */ 1079 netdev_features_t hw_features; 1080 /* user-requested features */ 1081 netdev_features_t wanted_features; 1082 /* mask of features inheritable by VLAN devices */ 1083 netdev_features_t vlan_features; 1084 1085 /* Interface index. Unique device identifier */ 1086 int ifindex; 1087 int iflink; 1088 1089 struct net_device_stats stats; 1090 atomic_long_t rx_dropped; /* dropped packets by core network 1091 * Do not use this in drivers. 1092 */ 1093 1094 #ifdef CONFIG_WIRELESS_EXT 1095 /* List of functions to handle Wireless Extensions (instead of ioctl). 1096 * See <net/iw_handler.h> for details. Jean II */ 1097 const struct iw_handler_def * wireless_handlers; 1098 /* Instance data managed by the core of Wireless Extensions. */ 1099 struct iw_public_data * wireless_data; 1100 #endif 1101 /* Management operations */ 1102 const struct net_device_ops *netdev_ops; 1103 const struct ethtool_ops *ethtool_ops; 1104 1105 /* Hardware header description */ 1106 const struct header_ops *header_ops; 1107 1108 unsigned int flags; /* interface flags (a la BSD) */ 1109 unsigned int priv_flags; /* Like 'flags' but invisible to userspace. 1110 * See if.h for definitions. */ 1111 unsigned short gflags; 1112 unsigned short padded; /* How much padding added by alloc_netdev() */ 1113 1114 unsigned char operstate; /* RFC2863 operstate */ 1115 unsigned char link_mode; /* mapping policy to operstate */ 1116 1117 unsigned char if_port; /* Selectable AUI, TP,..*/ 1118 unsigned char dma; /* DMA channel */ 1119 1120 unsigned int mtu; /* interface MTU value */ 1121 unsigned short type; /* interface hardware type */ 1122 unsigned short hard_header_len; /* hardware hdr length */ 1123 1124 /* extra head- and tailroom the hardware may need, but not in all cases 1125 * can this be guaranteed, especially tailroom. Some cases also use 1126 * LL_MAX_HEADER instead to allocate the skb. 1127 */ 1128 unsigned short needed_headroom; 1129 unsigned short needed_tailroom; 1130 1131 /* Interface address info. */ 1132 unsigned char perm_addr[MAX_ADDR_LEN]; /* permanent hw address */ 1133 unsigned char addr_assign_type; /* hw address assignment type */ 1134 unsigned char addr_len; /* hardware address length */ 1135 unsigned char neigh_priv_len; 1136 unsigned short dev_id; /* for shared network cards */ 1137 1138 spinlock_t addr_list_lock; 1139 struct netdev_hw_addr_list uc; /* Unicast mac addresses */ 1140 struct netdev_hw_addr_list mc; /* Multicast mac addresses */ 1141 bool uc_promisc; 1142 unsigned int promiscuity; 1143 unsigned int allmulti; 1144 1145 1146 /* Protocol specific pointers */ 1147 1148 #if IS_ENABLED(CONFIG_VLAN_8021Q) 1149 struct vlan_info __rcu *vlan_info; /* VLAN info */ 1150 #endif 1151 #if IS_ENABLED(CONFIG_NET_DSA) 1152 struct dsa_switch_tree *dsa_ptr; /* dsa specific data */ 1153 #endif 1154 void *atalk_ptr; /* AppleTalk link */ 1155 struct in_device __rcu *ip_ptr; /* IPv4 specific data */ 1156 struct dn_dev __rcu *dn_ptr; /* DECnet specific data */ 1157 struct inet6_dev __rcu *ip6_ptr; /* IPv6 specific data */ 1158 void *ax25_ptr; /* AX.25 specific data */ 1159 struct wireless_dev *ieee80211_ptr; /* IEEE 802.11 specific data, 1160 assign before registering */ 1161 1162 /* 1163 * Cache lines mostly used on receive path (including eth_type_trans()) 1164 */ 1165 unsigned long last_rx; /* Time of last Rx 1166 * This should not be set in 1167 * drivers, unless really needed, 1168 * because network stack (bonding) 1169 * use it if/when necessary, to 1170 * avoid dirtying this cache line. 1171 */ 1172 1173 struct net_device *master; /* Pointer to master device of a group, 1174 * which this device is member of. 1175 */ 1176 1177 /* Interface address info used in eth_type_trans() */ 1178 unsigned char *dev_addr; /* hw address, (before bcast 1179 because most packets are 1180 unicast) */ 1181 1182 struct netdev_hw_addr_list dev_addrs; /* list of device 1183 hw addresses */ 1184 1185 unsigned char broadcast[MAX_ADDR_LEN]; /* hw bcast add */ 1186 1187 #ifdef CONFIG_SYSFS 1188 struct kset *queues_kset; 1189 #endif 1190 1191 #ifdef CONFIG_RPS 1192 struct netdev_rx_queue *_rx; 1193 1194 /* Number of RX queues allocated at register_netdev() time */ 1195 unsigned int num_rx_queues; 1196 1197 /* Number of RX queues currently active in device */ 1198 unsigned int real_num_rx_queues; 1199 1200 #ifdef CONFIG_RFS_ACCEL 1201 /* CPU reverse-mapping for RX completion interrupts, indexed 1202 * by RX queue number. Assigned by driver. This must only be 1203 * set if the ndo_rx_flow_steer operation is defined. */ 1204 struct cpu_rmap *rx_cpu_rmap; 1205 #endif 1206 #endif 1207 1208 rx_handler_func_t __rcu *rx_handler; 1209 void __rcu *rx_handler_data; 1210 1211 struct netdev_queue __rcu *ingress_queue; 1212 1213 /* 1214 * Cache lines mostly used on transmit path 1215 */ 1216 struct netdev_queue *_tx ____cacheline_aligned_in_smp; 1217 1218 /* Number of TX queues allocated at alloc_netdev_mq() time */ 1219 unsigned int num_tx_queues; 1220 1221 /* Number of TX queues currently active in device */ 1222 unsigned int real_num_tx_queues; 1223 1224 /* root qdisc from userspace point of view */ 1225 struct Qdisc *qdisc; 1226 1227 unsigned long tx_queue_len; /* Max frames per queue allowed */ 1228 spinlock_t tx_global_lock; 1229 1230 #ifdef CONFIG_XPS 1231 struct xps_dev_maps __rcu *xps_maps; 1232 #endif 1233 1234 /* These may be needed for future network-power-down code. */ 1235 1236 /* 1237 * trans_start here is expensive for high speed devices on SMP, 1238 * please use netdev_queue->trans_start instead. 1239 */ 1240 unsigned long trans_start; /* Time (in jiffies) of last Tx */ 1241 1242 int watchdog_timeo; /* used by dev_watchdog() */ 1243 struct timer_list watchdog_timer; 1244 1245 /* Number of references to this device */ 1246 int __percpu *pcpu_refcnt; 1247 1248 /* delayed register/unregister */ 1249 struct list_head todo_list; 1250 /* device index hash chain */ 1251 struct hlist_node index_hlist; 1252 1253 struct list_head link_watch_list; 1254 1255 /* register/unregister state machine */ 1256 enum { NETREG_UNINITIALIZED=0, 1257 NETREG_REGISTERED, /* completed register_netdevice */ 1258 NETREG_UNREGISTERING, /* called unregister_netdevice */ 1259 NETREG_UNREGISTERED, /* completed unregister todo */ 1260 NETREG_RELEASED, /* called free_netdev */ 1261 NETREG_DUMMY, /* dummy device for NAPI poll */ 1262 } reg_state:8; 1263 1264 bool dismantle; /* device is going do be freed */ 1265 1266 enum { 1267 RTNL_LINK_INITIALIZED, 1268 RTNL_LINK_INITIALIZING, 1269 } rtnl_link_state:16; 1270 1271 /* Called from unregister, can be used to call free_netdev */ 1272 void (*destructor)(struct net_device *dev); 1273 1274 #ifdef CONFIG_NETPOLL 1275 struct netpoll_info *npinfo; 1276 #endif 1277 1278 #ifdef CONFIG_NET_NS 1279 /* Network namespace this network device is inside */ 1280 struct net *nd_net; 1281 #endif 1282 1283 /* mid-layer private */ 1284 union { 1285 void *ml_priv; 1286 struct pcpu_lstats __percpu *lstats; /* loopback stats */ 1287 struct pcpu_tstats __percpu *tstats; /* tunnel stats */ 1288 struct pcpu_dstats __percpu *dstats; /* dummy stats */ 1289 }; 1290 /* GARP */ 1291 struct garp_port __rcu *garp_port; 1292 1293 /* class/net/name entry */ 1294 struct device dev; 1295 /* space for optional device, statistics, and wireless sysfs groups */ 1296 const struct attribute_group *sysfs_groups[4]; 1297 1298 /* rtnetlink link ops */ 1299 const struct rtnl_link_ops *rtnl_link_ops; 1300 1301 /* for setting kernel sock attribute on TCP connection setup */ 1302 #define GSO_MAX_SIZE 65536 1303 unsigned int gso_max_size; 1304 #define GSO_MAX_SEGS 65535 1305 u16 gso_max_segs; 1306 1307 #ifdef CONFIG_DCB 1308 /* Data Center Bridging netlink ops */ 1309 const struct dcbnl_rtnl_ops *dcbnl_ops; 1310 #endif 1311 u8 num_tc; 1312 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE]; 1313 u8 prio_tc_map[TC_BITMASK + 1]; 1314 1315 #if IS_ENABLED(CONFIG_FCOE) 1316 /* max exchange id for FCoE LRO by ddp */ 1317 unsigned int fcoe_ddp_xid; 1318 #endif 1319 #if IS_ENABLED(CONFIG_NETPRIO_CGROUP) 1320 struct netprio_map __rcu *priomap; 1321 #endif 1322 /* phy device may attach itself for hardware timestamping */ 1323 struct phy_device *phydev; 1324 1325 struct lock_class_key *qdisc_tx_busylock; 1326 1327 /* group the device belongs to */ 1328 int group; 1329 1330 struct pm_qos_request pm_qos_req; 1331 }; 1332 #define to_net_dev(d) container_of(d, struct net_device, dev) 1333 1334 #define NETDEV_ALIGN 32 1335 1336 static inline 1337 int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio) 1338 { 1339 return dev->prio_tc_map[prio & TC_BITMASK]; 1340 } 1341 1342 static inline 1343 int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc) 1344 { 1345 if (tc >= dev->num_tc) 1346 return -EINVAL; 1347 1348 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK; 1349 return 0; 1350 } 1351 1352 static inline 1353 void netdev_reset_tc(struct net_device *dev) 1354 { 1355 dev->num_tc = 0; 1356 memset(dev->tc_to_txq, 0, sizeof(dev->tc_to_txq)); 1357 memset(dev->prio_tc_map, 0, sizeof(dev->prio_tc_map)); 1358 } 1359 1360 static inline 1361 int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset) 1362 { 1363 if (tc >= dev->num_tc) 1364 return -EINVAL; 1365 1366 dev->tc_to_txq[tc].count = count; 1367 dev->tc_to_txq[tc].offset = offset; 1368 return 0; 1369 } 1370 1371 static inline 1372 int netdev_set_num_tc(struct net_device *dev, u8 num_tc) 1373 { 1374 if (num_tc > TC_MAX_QUEUE) 1375 return -EINVAL; 1376 1377 dev->num_tc = num_tc; 1378 return 0; 1379 } 1380 1381 static inline 1382 int netdev_get_num_tc(struct net_device *dev) 1383 { 1384 return dev->num_tc; 1385 } 1386 1387 static inline 1388 struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev, 1389 unsigned int index) 1390 { 1391 return &dev->_tx[index]; 1392 } 1393 1394 static inline void netdev_for_each_tx_queue(struct net_device *dev, 1395 void (*f)(struct net_device *, 1396 struct netdev_queue *, 1397 void *), 1398 void *arg) 1399 { 1400 unsigned int i; 1401 1402 for (i = 0; i < dev->num_tx_queues; i++) 1403 f(dev, &dev->_tx[i], arg); 1404 } 1405 1406 extern struct netdev_queue *netdev_pick_tx(struct net_device *dev, 1407 struct sk_buff *skb); 1408 1409 /* 1410 * Net namespace inlines 1411 */ 1412 static inline 1413 struct net *dev_net(const struct net_device *dev) 1414 { 1415 return read_pnet(&dev->nd_net); 1416 } 1417 1418 static inline 1419 void dev_net_set(struct net_device *dev, struct net *net) 1420 { 1421 #ifdef CONFIG_NET_NS 1422 release_net(dev->nd_net); 1423 dev->nd_net = hold_net(net); 1424 #endif 1425 } 1426 1427 static inline bool netdev_uses_dsa_tags(struct net_device *dev) 1428 { 1429 #ifdef CONFIG_NET_DSA_TAG_DSA 1430 if (dev->dsa_ptr != NULL) 1431 return dsa_uses_dsa_tags(dev->dsa_ptr); 1432 #endif 1433 1434 return 0; 1435 } 1436 1437 static inline bool netdev_uses_trailer_tags(struct net_device *dev) 1438 { 1439 #ifdef CONFIG_NET_DSA_TAG_TRAILER 1440 if (dev->dsa_ptr != NULL) 1441 return dsa_uses_trailer_tags(dev->dsa_ptr); 1442 #endif 1443 1444 return 0; 1445 } 1446 1447 /** 1448 * netdev_priv - access network device private data 1449 * @dev: network device 1450 * 1451 * Get network device private data 1452 */ 1453 static inline void *netdev_priv(const struct net_device *dev) 1454 { 1455 return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN); 1456 } 1457 1458 /* Set the sysfs physical device reference for the network logical device 1459 * if set prior to registration will cause a symlink during initialization. 1460 */ 1461 #define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev)) 1462 1463 /* Set the sysfs device type for the network logical device to allow 1464 * fin grained indentification of different network device types. For 1465 * example Ethernet, Wirelss LAN, Bluetooth, WiMAX etc. 1466 */ 1467 #define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype)) 1468 1469 /** 1470 * netif_napi_add - initialize a napi context 1471 * @dev: network device 1472 * @napi: napi context 1473 * @poll: polling function 1474 * @weight: default weight 1475 * 1476 * netif_napi_add() must be used to initialize a napi context prior to calling 1477 * *any* of the other napi related functions. 1478 */ 1479 void netif_napi_add(struct net_device *dev, struct napi_struct *napi, 1480 int (*poll)(struct napi_struct *, int), int weight); 1481 1482 /** 1483 * netif_napi_del - remove a napi context 1484 * @napi: napi context 1485 * 1486 * netif_napi_del() removes a napi context from the network device napi list 1487 */ 1488 void netif_napi_del(struct napi_struct *napi); 1489 1490 struct napi_gro_cb { 1491 /* Virtual address of skb_shinfo(skb)->frags[0].page + offset. */ 1492 void *frag0; 1493 1494 /* Length of frag0. */ 1495 unsigned int frag0_len; 1496 1497 /* This indicates where we are processing relative to skb->data. */ 1498 int data_offset; 1499 1500 /* This is non-zero if the packet may be of the same flow. */ 1501 int same_flow; 1502 1503 /* This is non-zero if the packet cannot be merged with the new skb. */ 1504 int flush; 1505 1506 /* Number of segments aggregated. */ 1507 int count; 1508 1509 /* Free the skb? */ 1510 int free; 1511 #define NAPI_GRO_FREE 1 1512 #define NAPI_GRO_FREE_STOLEN_HEAD 2 1513 }; 1514 1515 #define NAPI_GRO_CB(skb) ((struct napi_gro_cb *)(skb)->cb) 1516 1517 struct packet_type { 1518 __be16 type; /* This is really htons(ether_type). */ 1519 struct net_device *dev; /* NULL is wildcarded here */ 1520 int (*func) (struct sk_buff *, 1521 struct net_device *, 1522 struct packet_type *, 1523 struct net_device *); 1524 struct sk_buff *(*gso_segment)(struct sk_buff *skb, 1525 netdev_features_t features); 1526 int (*gso_send_check)(struct sk_buff *skb); 1527 struct sk_buff **(*gro_receive)(struct sk_buff **head, 1528 struct sk_buff *skb); 1529 int (*gro_complete)(struct sk_buff *skb); 1530 bool (*id_match)(struct packet_type *ptype, 1531 struct sock *sk); 1532 void *af_packet_priv; 1533 struct list_head list; 1534 }; 1535 1536 #include <linux/notifier.h> 1537 1538 /* netdevice notifier chain. Please remember to update the rtnetlink 1539 * notification exclusion list in rtnetlink_event() when adding new 1540 * types. 1541 */ 1542 #define NETDEV_UP 0x0001 /* For now you can't veto a device up/down */ 1543 #define NETDEV_DOWN 0x0002 1544 #define NETDEV_REBOOT 0x0003 /* Tell a protocol stack a network interface 1545 detected a hardware crash and restarted 1546 - we can use this eg to kick tcp sessions 1547 once done */ 1548 #define NETDEV_CHANGE 0x0004 /* Notify device state change */ 1549 #define NETDEV_REGISTER 0x0005 1550 #define NETDEV_UNREGISTER 0x0006 1551 #define NETDEV_CHANGEMTU 0x0007 1552 #define NETDEV_CHANGEADDR 0x0008 1553 #define NETDEV_GOING_DOWN 0x0009 1554 #define NETDEV_CHANGENAME 0x000A 1555 #define NETDEV_FEAT_CHANGE 0x000B 1556 #define NETDEV_BONDING_FAILOVER 0x000C 1557 #define NETDEV_PRE_UP 0x000D 1558 #define NETDEV_PRE_TYPE_CHANGE 0x000E 1559 #define NETDEV_POST_TYPE_CHANGE 0x000F 1560 #define NETDEV_POST_INIT 0x0010 1561 #define NETDEV_UNREGISTER_FINAL 0x0011 1562 #define NETDEV_RELEASE 0x0012 1563 #define NETDEV_NOTIFY_PEERS 0x0013 1564 #define NETDEV_JOIN 0x0014 1565 1566 extern int register_netdevice_notifier(struct notifier_block *nb); 1567 extern int unregister_netdevice_notifier(struct notifier_block *nb); 1568 extern int call_netdevice_notifiers(unsigned long val, struct net_device *dev); 1569 1570 1571 extern rwlock_t dev_base_lock; /* Device list lock */ 1572 1573 1574 #define for_each_netdev(net, d) \ 1575 list_for_each_entry(d, &(net)->dev_base_head, dev_list) 1576 #define for_each_netdev_reverse(net, d) \ 1577 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list) 1578 #define for_each_netdev_rcu(net, d) \ 1579 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list) 1580 #define for_each_netdev_safe(net, d, n) \ 1581 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list) 1582 #define for_each_netdev_continue(net, d) \ 1583 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list) 1584 #define for_each_netdev_continue_rcu(net, d) \ 1585 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list) 1586 #define net_device_entry(lh) list_entry(lh, struct net_device, dev_list) 1587 1588 static inline struct net_device *next_net_device(struct net_device *dev) 1589 { 1590 struct list_head *lh; 1591 struct net *net; 1592 1593 net = dev_net(dev); 1594 lh = dev->dev_list.next; 1595 return lh == &net->dev_base_head ? NULL : net_device_entry(lh); 1596 } 1597 1598 static inline struct net_device *next_net_device_rcu(struct net_device *dev) 1599 { 1600 struct list_head *lh; 1601 struct net *net; 1602 1603 net = dev_net(dev); 1604 lh = rcu_dereference(list_next_rcu(&dev->dev_list)); 1605 return lh == &net->dev_base_head ? NULL : net_device_entry(lh); 1606 } 1607 1608 static inline struct net_device *first_net_device(struct net *net) 1609 { 1610 return list_empty(&net->dev_base_head) ? NULL : 1611 net_device_entry(net->dev_base_head.next); 1612 } 1613 1614 static inline struct net_device *first_net_device_rcu(struct net *net) 1615 { 1616 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head)); 1617 1618 return lh == &net->dev_base_head ? NULL : net_device_entry(lh); 1619 } 1620 1621 extern int netdev_boot_setup_check(struct net_device *dev); 1622 extern unsigned long netdev_boot_base(const char *prefix, int unit); 1623 extern struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type, 1624 const char *hwaddr); 1625 extern struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type); 1626 extern struct net_device *__dev_getfirstbyhwtype(struct net *net, unsigned short type); 1627 extern void dev_add_pack(struct packet_type *pt); 1628 extern void dev_remove_pack(struct packet_type *pt); 1629 extern void __dev_remove_pack(struct packet_type *pt); 1630 1631 extern struct net_device *dev_get_by_flags_rcu(struct net *net, unsigned short flags, 1632 unsigned short mask); 1633 extern struct net_device *dev_get_by_name(struct net *net, const char *name); 1634 extern struct net_device *dev_get_by_name_rcu(struct net *net, const char *name); 1635 extern struct net_device *__dev_get_by_name(struct net *net, const char *name); 1636 extern int dev_alloc_name(struct net_device *dev, const char *name); 1637 extern int dev_open(struct net_device *dev); 1638 extern int dev_close(struct net_device *dev); 1639 extern void dev_disable_lro(struct net_device *dev); 1640 extern int dev_loopback_xmit(struct sk_buff *newskb); 1641 extern int dev_queue_xmit(struct sk_buff *skb); 1642 extern int register_netdevice(struct net_device *dev); 1643 extern void unregister_netdevice_queue(struct net_device *dev, 1644 struct list_head *head); 1645 extern void unregister_netdevice_many(struct list_head *head); 1646 static inline void unregister_netdevice(struct net_device *dev) 1647 { 1648 unregister_netdevice_queue(dev, NULL); 1649 } 1650 1651 extern int netdev_refcnt_read(const struct net_device *dev); 1652 extern void free_netdev(struct net_device *dev); 1653 extern void synchronize_net(void); 1654 extern int init_dummy_netdev(struct net_device *dev); 1655 extern void netdev_resync_ops(struct net_device *dev); 1656 1657 extern struct net_device *dev_get_by_index(struct net *net, int ifindex); 1658 extern struct net_device *__dev_get_by_index(struct net *net, int ifindex); 1659 extern struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex); 1660 extern int dev_restart(struct net_device *dev); 1661 #ifdef CONFIG_NETPOLL_TRAP 1662 extern int netpoll_trap(void); 1663 #endif 1664 extern int skb_gro_receive(struct sk_buff **head, 1665 struct sk_buff *skb); 1666 extern void skb_gro_reset_offset(struct sk_buff *skb); 1667 1668 static inline unsigned int skb_gro_offset(const struct sk_buff *skb) 1669 { 1670 return NAPI_GRO_CB(skb)->data_offset; 1671 } 1672 1673 static inline unsigned int skb_gro_len(const struct sk_buff *skb) 1674 { 1675 return skb->len - NAPI_GRO_CB(skb)->data_offset; 1676 } 1677 1678 static inline void skb_gro_pull(struct sk_buff *skb, unsigned int len) 1679 { 1680 NAPI_GRO_CB(skb)->data_offset += len; 1681 } 1682 1683 static inline void *skb_gro_header_fast(struct sk_buff *skb, 1684 unsigned int offset) 1685 { 1686 return NAPI_GRO_CB(skb)->frag0 + offset; 1687 } 1688 1689 static inline int skb_gro_header_hard(struct sk_buff *skb, unsigned int hlen) 1690 { 1691 return NAPI_GRO_CB(skb)->frag0_len < hlen; 1692 } 1693 1694 static inline void *skb_gro_header_slow(struct sk_buff *skb, unsigned int hlen, 1695 unsigned int offset) 1696 { 1697 if (!pskb_may_pull(skb, hlen)) 1698 return NULL; 1699 1700 NAPI_GRO_CB(skb)->frag0 = NULL; 1701 NAPI_GRO_CB(skb)->frag0_len = 0; 1702 return skb->data + offset; 1703 } 1704 1705 static inline void *skb_gro_mac_header(struct sk_buff *skb) 1706 { 1707 return NAPI_GRO_CB(skb)->frag0 ?: skb_mac_header(skb); 1708 } 1709 1710 static inline void *skb_gro_network_header(struct sk_buff *skb) 1711 { 1712 return (NAPI_GRO_CB(skb)->frag0 ?: skb->data) + 1713 skb_network_offset(skb); 1714 } 1715 1716 static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev, 1717 unsigned short type, 1718 const void *daddr, const void *saddr, 1719 unsigned int len) 1720 { 1721 if (!dev->header_ops || !dev->header_ops->create) 1722 return 0; 1723 1724 return dev->header_ops->create(skb, dev, type, daddr, saddr, len); 1725 } 1726 1727 static inline int dev_parse_header(const struct sk_buff *skb, 1728 unsigned char *haddr) 1729 { 1730 const struct net_device *dev = skb->dev; 1731 1732 if (!dev->header_ops || !dev->header_ops->parse) 1733 return 0; 1734 return dev->header_ops->parse(skb, haddr); 1735 } 1736 1737 typedef int gifconf_func_t(struct net_device * dev, char __user * bufptr, int len); 1738 extern int register_gifconf(unsigned int family, gifconf_func_t * gifconf); 1739 static inline int unregister_gifconf(unsigned int family) 1740 { 1741 return register_gifconf(family, NULL); 1742 } 1743 1744 /* 1745 * Incoming packets are placed on per-cpu queues 1746 */ 1747 struct softnet_data { 1748 struct Qdisc *output_queue; 1749 struct Qdisc **output_queue_tailp; 1750 struct list_head poll_list; 1751 struct sk_buff *completion_queue; 1752 struct sk_buff_head process_queue; 1753 1754 /* stats */ 1755 unsigned int processed; 1756 unsigned int time_squeeze; 1757 unsigned int cpu_collision; 1758 unsigned int received_rps; 1759 1760 #ifdef CONFIG_RPS 1761 struct softnet_data *rps_ipi_list; 1762 1763 /* Elements below can be accessed between CPUs for RPS */ 1764 struct call_single_data csd ____cacheline_aligned_in_smp; 1765 struct softnet_data *rps_ipi_next; 1766 unsigned int cpu; 1767 unsigned int input_queue_head; 1768 unsigned int input_queue_tail; 1769 #endif 1770 unsigned int dropped; 1771 struct sk_buff_head input_pkt_queue; 1772 struct napi_struct backlog; 1773 }; 1774 1775 static inline void input_queue_head_incr(struct softnet_data *sd) 1776 { 1777 #ifdef CONFIG_RPS 1778 sd->input_queue_head++; 1779 #endif 1780 } 1781 1782 static inline void input_queue_tail_incr_save(struct softnet_data *sd, 1783 unsigned int *qtail) 1784 { 1785 #ifdef CONFIG_RPS 1786 *qtail = ++sd->input_queue_tail; 1787 #endif 1788 } 1789 1790 DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data); 1791 1792 extern void __netif_schedule(struct Qdisc *q); 1793 1794 static inline void netif_schedule_queue(struct netdev_queue *txq) 1795 { 1796 if (!(txq->state & QUEUE_STATE_ANY_XOFF)) 1797 __netif_schedule(txq->qdisc); 1798 } 1799 1800 static inline void netif_tx_schedule_all(struct net_device *dev) 1801 { 1802 unsigned int i; 1803 1804 for (i = 0; i < dev->num_tx_queues; i++) 1805 netif_schedule_queue(netdev_get_tx_queue(dev, i)); 1806 } 1807 1808 static inline void netif_tx_start_queue(struct netdev_queue *dev_queue) 1809 { 1810 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state); 1811 } 1812 1813 /** 1814 * netif_start_queue - allow transmit 1815 * @dev: network device 1816 * 1817 * Allow upper layers to call the device hard_start_xmit routine. 1818 */ 1819 static inline void netif_start_queue(struct net_device *dev) 1820 { 1821 netif_tx_start_queue(netdev_get_tx_queue(dev, 0)); 1822 } 1823 1824 static inline void netif_tx_start_all_queues(struct net_device *dev) 1825 { 1826 unsigned int i; 1827 1828 for (i = 0; i < dev->num_tx_queues; i++) { 1829 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 1830 netif_tx_start_queue(txq); 1831 } 1832 } 1833 1834 static inline void netif_tx_wake_queue(struct netdev_queue *dev_queue) 1835 { 1836 #ifdef CONFIG_NETPOLL_TRAP 1837 if (netpoll_trap()) { 1838 netif_tx_start_queue(dev_queue); 1839 return; 1840 } 1841 #endif 1842 if (test_and_clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state)) 1843 __netif_schedule(dev_queue->qdisc); 1844 } 1845 1846 /** 1847 * netif_wake_queue - restart transmit 1848 * @dev: network device 1849 * 1850 * Allow upper layers to call the device hard_start_xmit routine. 1851 * Used for flow control when transmit resources are available. 1852 */ 1853 static inline void netif_wake_queue(struct net_device *dev) 1854 { 1855 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0)); 1856 } 1857 1858 static inline void netif_tx_wake_all_queues(struct net_device *dev) 1859 { 1860 unsigned int i; 1861 1862 for (i = 0; i < dev->num_tx_queues; i++) { 1863 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 1864 netif_tx_wake_queue(txq); 1865 } 1866 } 1867 1868 static inline void netif_tx_stop_queue(struct netdev_queue *dev_queue) 1869 { 1870 if (WARN_ON(!dev_queue)) { 1871 pr_info("netif_stop_queue() cannot be called before register_netdev()\n"); 1872 return; 1873 } 1874 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state); 1875 } 1876 1877 /** 1878 * netif_stop_queue - stop transmitted packets 1879 * @dev: network device 1880 * 1881 * Stop upper layers calling the device hard_start_xmit routine. 1882 * Used for flow control when transmit resources are unavailable. 1883 */ 1884 static inline void netif_stop_queue(struct net_device *dev) 1885 { 1886 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0)); 1887 } 1888 1889 static inline void netif_tx_stop_all_queues(struct net_device *dev) 1890 { 1891 unsigned int i; 1892 1893 for (i = 0; i < dev->num_tx_queues; i++) { 1894 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 1895 netif_tx_stop_queue(txq); 1896 } 1897 } 1898 1899 static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue) 1900 { 1901 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state); 1902 } 1903 1904 /** 1905 * netif_queue_stopped - test if transmit queue is flowblocked 1906 * @dev: network device 1907 * 1908 * Test if transmit queue on device is currently unable to send. 1909 */ 1910 static inline bool netif_queue_stopped(const struct net_device *dev) 1911 { 1912 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0)); 1913 } 1914 1915 static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue) 1916 { 1917 return dev_queue->state & QUEUE_STATE_ANY_XOFF; 1918 } 1919 1920 static inline bool netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue) 1921 { 1922 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN; 1923 } 1924 1925 static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue, 1926 unsigned int bytes) 1927 { 1928 #ifdef CONFIG_BQL 1929 dql_queued(&dev_queue->dql, bytes); 1930 1931 if (likely(dql_avail(&dev_queue->dql) >= 0)) 1932 return; 1933 1934 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state); 1935 1936 /* 1937 * The XOFF flag must be set before checking the dql_avail below, 1938 * because in netdev_tx_completed_queue we update the dql_completed 1939 * before checking the XOFF flag. 1940 */ 1941 smp_mb(); 1942 1943 /* check again in case another CPU has just made room avail */ 1944 if (unlikely(dql_avail(&dev_queue->dql) >= 0)) 1945 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state); 1946 #endif 1947 } 1948 1949 static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes) 1950 { 1951 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes); 1952 } 1953 1954 static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue, 1955 unsigned int pkts, unsigned int bytes) 1956 { 1957 #ifdef CONFIG_BQL 1958 if (unlikely(!bytes)) 1959 return; 1960 1961 dql_completed(&dev_queue->dql, bytes); 1962 1963 /* 1964 * Without the memory barrier there is a small possiblity that 1965 * netdev_tx_sent_queue will miss the update and cause the queue to 1966 * be stopped forever 1967 */ 1968 smp_mb(); 1969 1970 if (dql_avail(&dev_queue->dql) < 0) 1971 return; 1972 1973 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state)) 1974 netif_schedule_queue(dev_queue); 1975 #endif 1976 } 1977 1978 static inline void netdev_completed_queue(struct net_device *dev, 1979 unsigned int pkts, unsigned int bytes) 1980 { 1981 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes); 1982 } 1983 1984 static inline void netdev_tx_reset_queue(struct netdev_queue *q) 1985 { 1986 #ifdef CONFIG_BQL 1987 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state); 1988 dql_reset(&q->dql); 1989 #endif 1990 } 1991 1992 static inline void netdev_reset_queue(struct net_device *dev_queue) 1993 { 1994 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0)); 1995 } 1996 1997 /** 1998 * netif_running - test if up 1999 * @dev: network device 2000 * 2001 * Test if the device has been brought up. 2002 */ 2003 static inline bool netif_running(const struct net_device *dev) 2004 { 2005 return test_bit(__LINK_STATE_START, &dev->state); 2006 } 2007 2008 /* 2009 * Routines to manage the subqueues on a device. We only need start 2010 * stop, and a check if it's stopped. All other device management is 2011 * done at the overall netdevice level. 2012 * Also test the device if we're multiqueue. 2013 */ 2014 2015 /** 2016 * netif_start_subqueue - allow sending packets on subqueue 2017 * @dev: network device 2018 * @queue_index: sub queue index 2019 * 2020 * Start individual transmit queue of a device with multiple transmit queues. 2021 */ 2022 static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index) 2023 { 2024 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 2025 2026 netif_tx_start_queue(txq); 2027 } 2028 2029 /** 2030 * netif_stop_subqueue - stop sending packets on subqueue 2031 * @dev: network device 2032 * @queue_index: sub queue index 2033 * 2034 * Stop individual transmit queue of a device with multiple transmit queues. 2035 */ 2036 static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index) 2037 { 2038 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 2039 #ifdef CONFIG_NETPOLL_TRAP 2040 if (netpoll_trap()) 2041 return; 2042 #endif 2043 netif_tx_stop_queue(txq); 2044 } 2045 2046 /** 2047 * netif_subqueue_stopped - test status of subqueue 2048 * @dev: network device 2049 * @queue_index: sub queue index 2050 * 2051 * Check individual transmit queue of a device with multiple transmit queues. 2052 */ 2053 static inline bool __netif_subqueue_stopped(const struct net_device *dev, 2054 u16 queue_index) 2055 { 2056 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 2057 2058 return netif_tx_queue_stopped(txq); 2059 } 2060 2061 static inline bool netif_subqueue_stopped(const struct net_device *dev, 2062 struct sk_buff *skb) 2063 { 2064 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb)); 2065 } 2066 2067 /** 2068 * netif_wake_subqueue - allow sending packets on subqueue 2069 * @dev: network device 2070 * @queue_index: sub queue index 2071 * 2072 * Resume individual transmit queue of a device with multiple transmit queues. 2073 */ 2074 static inline void netif_wake_subqueue(struct net_device *dev, u16 queue_index) 2075 { 2076 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 2077 #ifdef CONFIG_NETPOLL_TRAP 2078 if (netpoll_trap()) 2079 return; 2080 #endif 2081 if (test_and_clear_bit(__QUEUE_STATE_DRV_XOFF, &txq->state)) 2082 __netif_schedule(txq->qdisc); 2083 } 2084 2085 /* 2086 * Returns a Tx hash for the given packet when dev->real_num_tx_queues is used 2087 * as a distribution range limit for the returned value. 2088 */ 2089 static inline u16 skb_tx_hash(const struct net_device *dev, 2090 const struct sk_buff *skb) 2091 { 2092 return __skb_tx_hash(dev, skb, dev->real_num_tx_queues); 2093 } 2094 2095 /** 2096 * netif_is_multiqueue - test if device has multiple transmit queues 2097 * @dev: network device 2098 * 2099 * Check if device has multiple transmit queues 2100 */ 2101 static inline bool netif_is_multiqueue(const struct net_device *dev) 2102 { 2103 return dev->num_tx_queues > 1; 2104 } 2105 2106 extern int netif_set_real_num_tx_queues(struct net_device *dev, 2107 unsigned int txq); 2108 2109 #ifdef CONFIG_RPS 2110 extern int netif_set_real_num_rx_queues(struct net_device *dev, 2111 unsigned int rxq); 2112 #else 2113 static inline int netif_set_real_num_rx_queues(struct net_device *dev, 2114 unsigned int rxq) 2115 { 2116 return 0; 2117 } 2118 #endif 2119 2120 static inline int netif_copy_real_num_queues(struct net_device *to_dev, 2121 const struct net_device *from_dev) 2122 { 2123 int err; 2124 2125 err = netif_set_real_num_tx_queues(to_dev, 2126 from_dev->real_num_tx_queues); 2127 if (err) 2128 return err; 2129 #ifdef CONFIG_RPS 2130 return netif_set_real_num_rx_queues(to_dev, 2131 from_dev->real_num_rx_queues); 2132 #else 2133 return 0; 2134 #endif 2135 } 2136 2137 #define DEFAULT_MAX_NUM_RSS_QUEUES (8) 2138 extern int netif_get_num_default_rss_queues(void); 2139 2140 /* Use this variant when it is known for sure that it 2141 * is executing from hardware interrupt context or with hardware interrupts 2142 * disabled. 2143 */ 2144 extern void dev_kfree_skb_irq(struct sk_buff *skb); 2145 2146 /* Use this variant in places where it could be invoked 2147 * from either hardware interrupt or other context, with hardware interrupts 2148 * either disabled or enabled. 2149 */ 2150 extern void dev_kfree_skb_any(struct sk_buff *skb); 2151 2152 extern int netif_rx(struct sk_buff *skb); 2153 extern int netif_rx_ni(struct sk_buff *skb); 2154 extern int netif_receive_skb(struct sk_buff *skb); 2155 extern gro_result_t dev_gro_receive(struct napi_struct *napi, 2156 struct sk_buff *skb); 2157 extern gro_result_t napi_skb_finish(gro_result_t ret, struct sk_buff *skb); 2158 extern gro_result_t napi_gro_receive(struct napi_struct *napi, 2159 struct sk_buff *skb); 2160 extern void napi_gro_flush(struct napi_struct *napi); 2161 extern struct sk_buff * napi_get_frags(struct napi_struct *napi); 2162 extern gro_result_t napi_frags_finish(struct napi_struct *napi, 2163 struct sk_buff *skb, 2164 gro_result_t ret); 2165 extern gro_result_t napi_gro_frags(struct napi_struct *napi); 2166 2167 static inline void napi_free_frags(struct napi_struct *napi) 2168 { 2169 kfree_skb(napi->skb); 2170 napi->skb = NULL; 2171 } 2172 2173 extern int netdev_rx_handler_register(struct net_device *dev, 2174 rx_handler_func_t *rx_handler, 2175 void *rx_handler_data); 2176 extern void netdev_rx_handler_unregister(struct net_device *dev); 2177 2178 extern bool dev_valid_name(const char *name); 2179 extern int dev_ioctl(struct net *net, unsigned int cmd, void __user *); 2180 extern int dev_ethtool(struct net *net, struct ifreq *); 2181 extern unsigned int dev_get_flags(const struct net_device *); 2182 extern int __dev_change_flags(struct net_device *, unsigned int flags); 2183 extern int dev_change_flags(struct net_device *, unsigned int); 2184 extern void __dev_notify_flags(struct net_device *, unsigned int old_flags); 2185 extern int dev_change_name(struct net_device *, const char *); 2186 extern int dev_set_alias(struct net_device *, const char *, size_t); 2187 extern int dev_change_net_namespace(struct net_device *, 2188 struct net *, const char *); 2189 extern int dev_set_mtu(struct net_device *, int); 2190 extern void dev_set_group(struct net_device *, int); 2191 extern int dev_set_mac_address(struct net_device *, 2192 struct sockaddr *); 2193 extern int dev_hard_start_xmit(struct sk_buff *skb, 2194 struct net_device *dev, 2195 struct netdev_queue *txq); 2196 extern int dev_forward_skb(struct net_device *dev, 2197 struct sk_buff *skb); 2198 2199 extern int netdev_budget; 2200 2201 /* Called by rtnetlink.c:rtnl_unlock() */ 2202 extern void netdev_run_todo(void); 2203 2204 /** 2205 * dev_put - release reference to device 2206 * @dev: network device 2207 * 2208 * Release reference to device to allow it to be freed. 2209 */ 2210 static inline void dev_put(struct net_device *dev) 2211 { 2212 this_cpu_dec(*dev->pcpu_refcnt); 2213 } 2214 2215 /** 2216 * dev_hold - get reference to device 2217 * @dev: network device 2218 * 2219 * Hold reference to device to keep it from being freed. 2220 */ 2221 static inline void dev_hold(struct net_device *dev) 2222 { 2223 this_cpu_inc(*dev->pcpu_refcnt); 2224 } 2225 2226 /* Carrier loss detection, dial on demand. The functions netif_carrier_on 2227 * and _off may be called from IRQ context, but it is caller 2228 * who is responsible for serialization of these calls. 2229 * 2230 * The name carrier is inappropriate, these functions should really be 2231 * called netif_lowerlayer_*() because they represent the state of any 2232 * kind of lower layer not just hardware media. 2233 */ 2234 2235 extern void linkwatch_init_dev(struct net_device *dev); 2236 extern void linkwatch_fire_event(struct net_device *dev); 2237 extern void linkwatch_forget_dev(struct net_device *dev); 2238 2239 /** 2240 * netif_carrier_ok - test if carrier present 2241 * @dev: network device 2242 * 2243 * Check if carrier is present on device 2244 */ 2245 static inline bool netif_carrier_ok(const struct net_device *dev) 2246 { 2247 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state); 2248 } 2249 2250 extern unsigned long dev_trans_start(struct net_device *dev); 2251 2252 extern void __netdev_watchdog_up(struct net_device *dev); 2253 2254 extern void netif_carrier_on(struct net_device *dev); 2255 2256 extern void netif_carrier_off(struct net_device *dev); 2257 2258 /** 2259 * netif_dormant_on - mark device as dormant. 2260 * @dev: network device 2261 * 2262 * Mark device as dormant (as per RFC2863). 2263 * 2264 * The dormant state indicates that the relevant interface is not 2265 * actually in a condition to pass packets (i.e., it is not 'up') but is 2266 * in a "pending" state, waiting for some external event. For "on- 2267 * demand" interfaces, this new state identifies the situation where the 2268 * interface is waiting for events to place it in the up state. 2269 * 2270 */ 2271 static inline void netif_dormant_on(struct net_device *dev) 2272 { 2273 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state)) 2274 linkwatch_fire_event(dev); 2275 } 2276 2277 /** 2278 * netif_dormant_off - set device as not dormant. 2279 * @dev: network device 2280 * 2281 * Device is not in dormant state. 2282 */ 2283 static inline void netif_dormant_off(struct net_device *dev) 2284 { 2285 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state)) 2286 linkwatch_fire_event(dev); 2287 } 2288 2289 /** 2290 * netif_dormant - test if carrier present 2291 * @dev: network device 2292 * 2293 * Check if carrier is present on device 2294 */ 2295 static inline bool netif_dormant(const struct net_device *dev) 2296 { 2297 return test_bit(__LINK_STATE_DORMANT, &dev->state); 2298 } 2299 2300 2301 /** 2302 * netif_oper_up - test if device is operational 2303 * @dev: network device 2304 * 2305 * Check if carrier is operational 2306 */ 2307 static inline bool netif_oper_up(const struct net_device *dev) 2308 { 2309 return (dev->operstate == IF_OPER_UP || 2310 dev->operstate == IF_OPER_UNKNOWN /* backward compat */); 2311 } 2312 2313 /** 2314 * netif_device_present - is device available or removed 2315 * @dev: network device 2316 * 2317 * Check if device has not been removed from system. 2318 */ 2319 static inline bool netif_device_present(struct net_device *dev) 2320 { 2321 return test_bit(__LINK_STATE_PRESENT, &dev->state); 2322 } 2323 2324 extern void netif_device_detach(struct net_device *dev); 2325 2326 extern void netif_device_attach(struct net_device *dev); 2327 2328 /* 2329 * Network interface message level settings 2330 */ 2331 2332 enum { 2333 NETIF_MSG_DRV = 0x0001, 2334 NETIF_MSG_PROBE = 0x0002, 2335 NETIF_MSG_LINK = 0x0004, 2336 NETIF_MSG_TIMER = 0x0008, 2337 NETIF_MSG_IFDOWN = 0x0010, 2338 NETIF_MSG_IFUP = 0x0020, 2339 NETIF_MSG_RX_ERR = 0x0040, 2340 NETIF_MSG_TX_ERR = 0x0080, 2341 NETIF_MSG_TX_QUEUED = 0x0100, 2342 NETIF_MSG_INTR = 0x0200, 2343 NETIF_MSG_TX_DONE = 0x0400, 2344 NETIF_MSG_RX_STATUS = 0x0800, 2345 NETIF_MSG_PKTDATA = 0x1000, 2346 NETIF_MSG_HW = 0x2000, 2347 NETIF_MSG_WOL = 0x4000, 2348 }; 2349 2350 #define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV) 2351 #define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE) 2352 #define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK) 2353 #define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER) 2354 #define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN) 2355 #define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP) 2356 #define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR) 2357 #define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR) 2358 #define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED) 2359 #define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR) 2360 #define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE) 2361 #define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS) 2362 #define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA) 2363 #define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW) 2364 #define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL) 2365 2366 static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits) 2367 { 2368 /* use default */ 2369 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8)) 2370 return default_msg_enable_bits; 2371 if (debug_value == 0) /* no output */ 2372 return 0; 2373 /* set low N bits */ 2374 return (1 << debug_value) - 1; 2375 } 2376 2377 static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu) 2378 { 2379 spin_lock(&txq->_xmit_lock); 2380 txq->xmit_lock_owner = cpu; 2381 } 2382 2383 static inline void __netif_tx_lock_bh(struct netdev_queue *txq) 2384 { 2385 spin_lock_bh(&txq->_xmit_lock); 2386 txq->xmit_lock_owner = smp_processor_id(); 2387 } 2388 2389 static inline bool __netif_tx_trylock(struct netdev_queue *txq) 2390 { 2391 bool ok = spin_trylock(&txq->_xmit_lock); 2392 if (likely(ok)) 2393 txq->xmit_lock_owner = smp_processor_id(); 2394 return ok; 2395 } 2396 2397 static inline void __netif_tx_unlock(struct netdev_queue *txq) 2398 { 2399 txq->xmit_lock_owner = -1; 2400 spin_unlock(&txq->_xmit_lock); 2401 } 2402 2403 static inline void __netif_tx_unlock_bh(struct netdev_queue *txq) 2404 { 2405 txq->xmit_lock_owner = -1; 2406 spin_unlock_bh(&txq->_xmit_lock); 2407 } 2408 2409 static inline void txq_trans_update(struct netdev_queue *txq) 2410 { 2411 if (txq->xmit_lock_owner != -1) 2412 txq->trans_start = jiffies; 2413 } 2414 2415 /** 2416 * netif_tx_lock - grab network device transmit lock 2417 * @dev: network device 2418 * 2419 * Get network device transmit lock 2420 */ 2421 static inline void netif_tx_lock(struct net_device *dev) 2422 { 2423 unsigned int i; 2424 int cpu; 2425 2426 spin_lock(&dev->tx_global_lock); 2427 cpu = smp_processor_id(); 2428 for (i = 0; i < dev->num_tx_queues; i++) { 2429 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 2430 2431 /* We are the only thread of execution doing a 2432 * freeze, but we have to grab the _xmit_lock in 2433 * order to synchronize with threads which are in 2434 * the ->hard_start_xmit() handler and already 2435 * checked the frozen bit. 2436 */ 2437 __netif_tx_lock(txq, cpu); 2438 set_bit(__QUEUE_STATE_FROZEN, &txq->state); 2439 __netif_tx_unlock(txq); 2440 } 2441 } 2442 2443 static inline void netif_tx_lock_bh(struct net_device *dev) 2444 { 2445 local_bh_disable(); 2446 netif_tx_lock(dev); 2447 } 2448 2449 static inline void netif_tx_unlock(struct net_device *dev) 2450 { 2451 unsigned int i; 2452 2453 for (i = 0; i < dev->num_tx_queues; i++) { 2454 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 2455 2456 /* No need to grab the _xmit_lock here. If the 2457 * queue is not stopped for another reason, we 2458 * force a schedule. 2459 */ 2460 clear_bit(__QUEUE_STATE_FROZEN, &txq->state); 2461 netif_schedule_queue(txq); 2462 } 2463 spin_unlock(&dev->tx_global_lock); 2464 } 2465 2466 static inline void netif_tx_unlock_bh(struct net_device *dev) 2467 { 2468 netif_tx_unlock(dev); 2469 local_bh_enable(); 2470 } 2471 2472 #define HARD_TX_LOCK(dev, txq, cpu) { \ 2473 if ((dev->features & NETIF_F_LLTX) == 0) { \ 2474 __netif_tx_lock(txq, cpu); \ 2475 } \ 2476 } 2477 2478 #define HARD_TX_UNLOCK(dev, txq) { \ 2479 if ((dev->features & NETIF_F_LLTX) == 0) { \ 2480 __netif_tx_unlock(txq); \ 2481 } \ 2482 } 2483 2484 static inline void netif_tx_disable(struct net_device *dev) 2485 { 2486 unsigned int i; 2487 int cpu; 2488 2489 local_bh_disable(); 2490 cpu = smp_processor_id(); 2491 for (i = 0; i < dev->num_tx_queues; i++) { 2492 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 2493 2494 __netif_tx_lock(txq, cpu); 2495 netif_tx_stop_queue(txq); 2496 __netif_tx_unlock(txq); 2497 } 2498 local_bh_enable(); 2499 } 2500 2501 static inline void netif_addr_lock(struct net_device *dev) 2502 { 2503 spin_lock(&dev->addr_list_lock); 2504 } 2505 2506 static inline void netif_addr_lock_nested(struct net_device *dev) 2507 { 2508 spin_lock_nested(&dev->addr_list_lock, SINGLE_DEPTH_NESTING); 2509 } 2510 2511 static inline void netif_addr_lock_bh(struct net_device *dev) 2512 { 2513 spin_lock_bh(&dev->addr_list_lock); 2514 } 2515 2516 static inline void netif_addr_unlock(struct net_device *dev) 2517 { 2518 spin_unlock(&dev->addr_list_lock); 2519 } 2520 2521 static inline void netif_addr_unlock_bh(struct net_device *dev) 2522 { 2523 spin_unlock_bh(&dev->addr_list_lock); 2524 } 2525 2526 /* 2527 * dev_addrs walker. Should be used only for read access. Call with 2528 * rcu_read_lock held. 2529 */ 2530 #define for_each_dev_addr(dev, ha) \ 2531 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list) 2532 2533 /* These functions live elsewhere (drivers/net/net_init.c, but related) */ 2534 2535 extern void ether_setup(struct net_device *dev); 2536 2537 /* Support for loadable net-drivers */ 2538 extern struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name, 2539 void (*setup)(struct net_device *), 2540 unsigned int txqs, unsigned int rxqs); 2541 #define alloc_netdev(sizeof_priv, name, setup) \ 2542 alloc_netdev_mqs(sizeof_priv, name, setup, 1, 1) 2543 2544 #define alloc_netdev_mq(sizeof_priv, name, setup, count) \ 2545 alloc_netdev_mqs(sizeof_priv, name, setup, count, count) 2546 2547 extern int register_netdev(struct net_device *dev); 2548 extern void unregister_netdev(struct net_device *dev); 2549 2550 /* General hardware address lists handling functions */ 2551 extern int __hw_addr_add_multiple(struct netdev_hw_addr_list *to_list, 2552 struct netdev_hw_addr_list *from_list, 2553 int addr_len, unsigned char addr_type); 2554 extern void __hw_addr_del_multiple(struct netdev_hw_addr_list *to_list, 2555 struct netdev_hw_addr_list *from_list, 2556 int addr_len, unsigned char addr_type); 2557 extern int __hw_addr_sync(struct netdev_hw_addr_list *to_list, 2558 struct netdev_hw_addr_list *from_list, 2559 int addr_len); 2560 extern void __hw_addr_unsync(struct netdev_hw_addr_list *to_list, 2561 struct netdev_hw_addr_list *from_list, 2562 int addr_len); 2563 extern void __hw_addr_flush(struct netdev_hw_addr_list *list); 2564 extern void __hw_addr_init(struct netdev_hw_addr_list *list); 2565 2566 /* Functions used for device addresses handling */ 2567 extern int dev_addr_add(struct net_device *dev, const unsigned char *addr, 2568 unsigned char addr_type); 2569 extern int dev_addr_del(struct net_device *dev, const unsigned char *addr, 2570 unsigned char addr_type); 2571 extern int dev_addr_add_multiple(struct net_device *to_dev, 2572 struct net_device *from_dev, 2573 unsigned char addr_type); 2574 extern int dev_addr_del_multiple(struct net_device *to_dev, 2575 struct net_device *from_dev, 2576 unsigned char addr_type); 2577 extern void dev_addr_flush(struct net_device *dev); 2578 extern int dev_addr_init(struct net_device *dev); 2579 2580 /* Functions used for unicast addresses handling */ 2581 extern int dev_uc_add(struct net_device *dev, const unsigned char *addr); 2582 extern int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr); 2583 extern int dev_uc_del(struct net_device *dev, const unsigned char *addr); 2584 extern int dev_uc_sync(struct net_device *to, struct net_device *from); 2585 extern void dev_uc_unsync(struct net_device *to, struct net_device *from); 2586 extern void dev_uc_flush(struct net_device *dev); 2587 extern void dev_uc_init(struct net_device *dev); 2588 2589 /* Functions used for multicast addresses handling */ 2590 extern int dev_mc_add(struct net_device *dev, const unsigned char *addr); 2591 extern int dev_mc_add_global(struct net_device *dev, const unsigned char *addr); 2592 extern int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr); 2593 extern int dev_mc_del(struct net_device *dev, const unsigned char *addr); 2594 extern int dev_mc_del_global(struct net_device *dev, const unsigned char *addr); 2595 extern int dev_mc_sync(struct net_device *to, struct net_device *from); 2596 extern void dev_mc_unsync(struct net_device *to, struct net_device *from); 2597 extern void dev_mc_flush(struct net_device *dev); 2598 extern void dev_mc_init(struct net_device *dev); 2599 2600 /* Functions used for secondary unicast and multicast support */ 2601 extern void dev_set_rx_mode(struct net_device *dev); 2602 extern void __dev_set_rx_mode(struct net_device *dev); 2603 extern int dev_set_promiscuity(struct net_device *dev, int inc); 2604 extern int dev_set_allmulti(struct net_device *dev, int inc); 2605 extern void netdev_state_change(struct net_device *dev); 2606 extern void netdev_notify_peers(struct net_device *dev); 2607 extern void netdev_features_change(struct net_device *dev); 2608 /* Load a device via the kmod */ 2609 extern void dev_load(struct net *net, const char *name); 2610 extern void dev_mcast_init(void); 2611 extern struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev, 2612 struct rtnl_link_stats64 *storage); 2613 extern void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64, 2614 const struct net_device_stats *netdev_stats); 2615 2616 extern int netdev_max_backlog; 2617 extern int netdev_tstamp_prequeue; 2618 extern int weight_p; 2619 extern int bpf_jit_enable; 2620 extern int netdev_set_master(struct net_device *dev, struct net_device *master); 2621 extern int netdev_set_bond_master(struct net_device *dev, 2622 struct net_device *master); 2623 extern int skb_checksum_help(struct sk_buff *skb); 2624 extern struct sk_buff *skb_gso_segment(struct sk_buff *skb, 2625 netdev_features_t features); 2626 #ifdef CONFIG_BUG 2627 extern void netdev_rx_csum_fault(struct net_device *dev); 2628 #else 2629 static inline void netdev_rx_csum_fault(struct net_device *dev) 2630 { 2631 } 2632 #endif 2633 /* rx skb timestamps */ 2634 extern void net_enable_timestamp(void); 2635 extern void net_disable_timestamp(void); 2636 2637 #ifdef CONFIG_PROC_FS 2638 extern void *dev_seq_start(struct seq_file *seq, loff_t *pos); 2639 extern void *dev_seq_next(struct seq_file *seq, void *v, loff_t *pos); 2640 extern void dev_seq_stop(struct seq_file *seq, void *v); 2641 #endif 2642 2643 extern int netdev_class_create_file(struct class_attribute *class_attr); 2644 extern void netdev_class_remove_file(struct class_attribute *class_attr); 2645 2646 extern struct kobj_ns_type_operations net_ns_type_operations; 2647 2648 extern const char *netdev_drivername(const struct net_device *dev); 2649 2650 extern void linkwatch_run_queue(void); 2651 2652 static inline netdev_features_t netdev_get_wanted_features( 2653 struct net_device *dev) 2654 { 2655 return (dev->features & ~dev->hw_features) | dev->wanted_features; 2656 } 2657 netdev_features_t netdev_increment_features(netdev_features_t all, 2658 netdev_features_t one, netdev_features_t mask); 2659 int __netdev_update_features(struct net_device *dev); 2660 void netdev_update_features(struct net_device *dev); 2661 void netdev_change_features(struct net_device *dev); 2662 2663 void netif_stacked_transfer_operstate(const struct net_device *rootdev, 2664 struct net_device *dev); 2665 2666 netdev_features_t netif_skb_features(struct sk_buff *skb); 2667 2668 static inline bool net_gso_ok(netdev_features_t features, int gso_type) 2669 { 2670 netdev_features_t feature = gso_type << NETIF_F_GSO_SHIFT; 2671 2672 /* check flags correspondence */ 2673 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT)); 2674 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_UFO >> NETIF_F_GSO_SHIFT)); 2675 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT)); 2676 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT)); 2677 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT)); 2678 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT)); 2679 2680 return (features & feature) == feature; 2681 } 2682 2683 static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features) 2684 { 2685 return net_gso_ok(features, skb_shinfo(skb)->gso_type) && 2686 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST)); 2687 } 2688 2689 static inline bool netif_needs_gso(struct sk_buff *skb, 2690 netdev_features_t features) 2691 { 2692 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) || 2693 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) && 2694 (skb->ip_summed != CHECKSUM_UNNECESSARY))); 2695 } 2696 2697 static inline void netif_set_gso_max_size(struct net_device *dev, 2698 unsigned int size) 2699 { 2700 dev->gso_max_size = size; 2701 } 2702 2703 static inline bool netif_is_bond_slave(struct net_device *dev) 2704 { 2705 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING; 2706 } 2707 2708 static inline bool netif_supports_nofcs(struct net_device *dev) 2709 { 2710 return dev->priv_flags & IFF_SUPP_NOFCS; 2711 } 2712 2713 extern struct pernet_operations __net_initdata loopback_net_ops; 2714 2715 /* Logging, debugging and troubleshooting/diagnostic helpers. */ 2716 2717 /* netdev_printk helpers, similar to dev_printk */ 2718 2719 static inline const char *netdev_name(const struct net_device *dev) 2720 { 2721 if (dev->reg_state != NETREG_REGISTERED) 2722 return "(unregistered net_device)"; 2723 return dev->name; 2724 } 2725 2726 extern __printf(3, 4) 2727 int netdev_printk(const char *level, const struct net_device *dev, 2728 const char *format, ...); 2729 extern __printf(2, 3) 2730 int netdev_emerg(const struct net_device *dev, const char *format, ...); 2731 extern __printf(2, 3) 2732 int netdev_alert(const struct net_device *dev, const char *format, ...); 2733 extern __printf(2, 3) 2734 int netdev_crit(const struct net_device *dev, const char *format, ...); 2735 extern __printf(2, 3) 2736 int netdev_err(const struct net_device *dev, const char *format, ...); 2737 extern __printf(2, 3) 2738 int netdev_warn(const struct net_device *dev, const char *format, ...); 2739 extern __printf(2, 3) 2740 int netdev_notice(const struct net_device *dev, const char *format, ...); 2741 extern __printf(2, 3) 2742 int netdev_info(const struct net_device *dev, const char *format, ...); 2743 2744 #define MODULE_ALIAS_NETDEV(device) \ 2745 MODULE_ALIAS("netdev-" device) 2746 2747 #if defined(CONFIG_DYNAMIC_DEBUG) 2748 #define netdev_dbg(__dev, format, args...) \ 2749 do { \ 2750 dynamic_netdev_dbg(__dev, format, ##args); \ 2751 } while (0) 2752 #elif defined(DEBUG) 2753 #define netdev_dbg(__dev, format, args...) \ 2754 netdev_printk(KERN_DEBUG, __dev, format, ##args) 2755 #else 2756 #define netdev_dbg(__dev, format, args...) \ 2757 ({ \ 2758 if (0) \ 2759 netdev_printk(KERN_DEBUG, __dev, format, ##args); \ 2760 0; \ 2761 }) 2762 #endif 2763 2764 #if defined(VERBOSE_DEBUG) 2765 #define netdev_vdbg netdev_dbg 2766 #else 2767 2768 #define netdev_vdbg(dev, format, args...) \ 2769 ({ \ 2770 if (0) \ 2771 netdev_printk(KERN_DEBUG, dev, format, ##args); \ 2772 0; \ 2773 }) 2774 #endif 2775 2776 /* 2777 * netdev_WARN() acts like dev_printk(), but with the key difference 2778 * of using a WARN/WARN_ON to get the message out, including the 2779 * file/line information and a backtrace. 2780 */ 2781 #define netdev_WARN(dev, format, args...) \ 2782 WARN(1, "netdevice: %s\n" format, netdev_name(dev), ##args); 2783 2784 /* netif printk helpers, similar to netdev_printk */ 2785 2786 #define netif_printk(priv, type, level, dev, fmt, args...) \ 2787 do { \ 2788 if (netif_msg_##type(priv)) \ 2789 netdev_printk(level, (dev), fmt, ##args); \ 2790 } while (0) 2791 2792 #define netif_level(level, priv, type, dev, fmt, args...) \ 2793 do { \ 2794 if (netif_msg_##type(priv)) \ 2795 netdev_##level(dev, fmt, ##args); \ 2796 } while (0) 2797 2798 #define netif_emerg(priv, type, dev, fmt, args...) \ 2799 netif_level(emerg, priv, type, dev, fmt, ##args) 2800 #define netif_alert(priv, type, dev, fmt, args...) \ 2801 netif_level(alert, priv, type, dev, fmt, ##args) 2802 #define netif_crit(priv, type, dev, fmt, args...) \ 2803 netif_level(crit, priv, type, dev, fmt, ##args) 2804 #define netif_err(priv, type, dev, fmt, args...) \ 2805 netif_level(err, priv, type, dev, fmt, ##args) 2806 #define netif_warn(priv, type, dev, fmt, args...) \ 2807 netif_level(warn, priv, type, dev, fmt, ##args) 2808 #define netif_notice(priv, type, dev, fmt, args...) \ 2809 netif_level(notice, priv, type, dev, fmt, ##args) 2810 #define netif_info(priv, type, dev, fmt, args...) \ 2811 netif_level(info, priv, type, dev, fmt, ##args) 2812 2813 #if defined(CONFIG_DYNAMIC_DEBUG) 2814 #define netif_dbg(priv, type, netdev, format, args...) \ 2815 do { \ 2816 if (netif_msg_##type(priv)) \ 2817 dynamic_netdev_dbg(netdev, format, ##args); \ 2818 } while (0) 2819 #elif defined(DEBUG) 2820 #define netif_dbg(priv, type, dev, format, args...) \ 2821 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args) 2822 #else 2823 #define netif_dbg(priv, type, dev, format, args...) \ 2824 ({ \ 2825 if (0) \ 2826 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \ 2827 0; \ 2828 }) 2829 #endif 2830 2831 #if defined(VERBOSE_DEBUG) 2832 #define netif_vdbg netif_dbg 2833 #else 2834 #define netif_vdbg(priv, type, dev, format, args...) \ 2835 ({ \ 2836 if (0) \ 2837 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \ 2838 0; \ 2839 }) 2840 #endif 2841 2842 #endif /* __KERNEL__ */ 2843 2844 #endif /* _LINUX_NETDEVICE_H */ 2845