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