1 /* SPDX-License-Identifier: GPL-2.0-or-later */ 2 /* 3 * INET An implementation of the TCP/IP protocol suite for the LINUX 4 * operating system. INET is implemented using the BSD Socket 5 * interface as the means of communication with the user level. 6 * 7 * Definitions for the Interfaces handler. 8 * 9 * Version: @(#)dev.h 1.0.10 08/12/93 10 * 11 * Authors: Ross Biro 12 * Fred N. van Kempen, <[email protected]> 13 * Corey Minyard <[email protected]> 14 * Donald J. Becker, <[email protected]> 15 * Alan Cox, <[email protected]> 16 * Bjorn Ekwall. <[email protected]> 17 * Pekka Riikonen <[email protected]> 18 * 19 * Moved to /usr/include/linux for NET3 20 */ 21 #ifndef _LINUX_NETDEVICE_H 22 #define _LINUX_NETDEVICE_H 23 24 #include <linux/timer.h> 25 #include <linux/bug.h> 26 #include <linux/delay.h> 27 #include <linux/atomic.h> 28 #include <linux/prefetch.h> 29 #include <asm/cache.h> 30 #include <asm/byteorder.h> 31 32 #include <linux/percpu.h> 33 #include <linux/rculist.h> 34 #include <linux/workqueue.h> 35 #include <linux/dynamic_queue_limits.h> 36 37 #include <net/net_namespace.h> 38 #ifdef CONFIG_DCB 39 #include <net/dcbnl.h> 40 #endif 41 #include <net/netprio_cgroup.h> 42 #include <net/xdp.h> 43 44 #include <linux/netdev_features.h> 45 #include <linux/neighbour.h> 46 #include <uapi/linux/netdevice.h> 47 #include <uapi/linux/if_bonding.h> 48 #include <uapi/linux/pkt_cls.h> 49 #include <linux/hashtable.h> 50 51 struct netpoll_info; 52 struct device; 53 struct ethtool_ops; 54 struct phy_device; 55 struct dsa_port; 56 struct ip_tunnel_parm; 57 struct macsec_context; 58 struct macsec_ops; 59 60 struct sfp_bus; 61 /* 802.11 specific */ 62 struct wireless_dev; 63 /* 802.15.4 specific */ 64 struct wpan_dev; 65 struct mpls_dev; 66 /* UDP Tunnel offloads */ 67 struct udp_tunnel_info; 68 struct udp_tunnel_nic_info; 69 struct udp_tunnel_nic; 70 struct bpf_prog; 71 struct xdp_buff; 72 73 void synchronize_net(void); 74 void netdev_set_default_ethtool_ops(struct net_device *dev, 75 const struct ethtool_ops *ops); 76 77 /* Backlog congestion levels */ 78 #define NET_RX_SUCCESS 0 /* keep 'em coming, baby */ 79 #define NET_RX_DROP 1 /* packet dropped */ 80 81 #define MAX_NEST_DEV 8 82 83 /* 84 * Transmit return codes: transmit return codes originate from three different 85 * namespaces: 86 * 87 * - qdisc return codes 88 * - driver transmit return codes 89 * - errno values 90 * 91 * Drivers are allowed to return any one of those in their hard_start_xmit() 92 * function. Real network devices commonly used with qdiscs should only return 93 * the driver transmit return codes though - when qdiscs are used, the actual 94 * transmission happens asynchronously, so the value is not propagated to 95 * higher layers. Virtual network devices transmit synchronously; in this case 96 * the driver transmit return codes are consumed by dev_queue_xmit(), and all 97 * others are propagated to higher layers. 98 */ 99 100 /* qdisc ->enqueue() return codes. */ 101 #define NET_XMIT_SUCCESS 0x00 102 #define NET_XMIT_DROP 0x01 /* skb dropped */ 103 #define NET_XMIT_CN 0x02 /* congestion notification */ 104 #define NET_XMIT_MASK 0x0f /* qdisc flags in net/sch_generic.h */ 105 106 /* NET_XMIT_CN is special. It does not guarantee that this packet is lost. It 107 * indicates that the device will soon be dropping packets, or already drops 108 * some packets of the same priority; prompting us to send less aggressively. */ 109 #define net_xmit_eval(e) ((e) == NET_XMIT_CN ? 0 : (e)) 110 #define net_xmit_errno(e) ((e) != NET_XMIT_CN ? -ENOBUFS : 0) 111 112 /* Driver transmit return codes */ 113 #define NETDEV_TX_MASK 0xf0 114 115 enum netdev_tx { 116 __NETDEV_TX_MIN = INT_MIN, /* make sure enum is signed */ 117 NETDEV_TX_OK = 0x00, /* driver took care of packet */ 118 NETDEV_TX_BUSY = 0x10, /* driver tx path was busy*/ 119 }; 120 typedef enum netdev_tx netdev_tx_t; 121 122 /* 123 * Current order: NETDEV_TX_MASK > NET_XMIT_MASK >= 0 is significant; 124 * hard_start_xmit() return < NET_XMIT_MASK means skb was consumed. 125 */ 126 static inline bool dev_xmit_complete(int rc) 127 { 128 /* 129 * Positive cases with an skb consumed by a driver: 130 * - successful transmission (rc == NETDEV_TX_OK) 131 * - error while transmitting (rc < 0) 132 * - error while queueing to a different device (rc & NET_XMIT_MASK) 133 */ 134 if (likely(rc < NET_XMIT_MASK)) 135 return true; 136 137 return false; 138 } 139 140 /* 141 * Compute the worst-case header length according to the protocols 142 * used. 143 */ 144 145 #if defined(CONFIG_HYPERV_NET) 146 # define LL_MAX_HEADER 128 147 #elif defined(CONFIG_WLAN) || IS_ENABLED(CONFIG_AX25) 148 # if defined(CONFIG_MAC80211_MESH) 149 # define LL_MAX_HEADER 128 150 # else 151 # define LL_MAX_HEADER 96 152 # endif 153 #else 154 # define LL_MAX_HEADER 32 155 #endif 156 157 #if !IS_ENABLED(CONFIG_NET_IPIP) && !IS_ENABLED(CONFIG_NET_IPGRE) && \ 158 !IS_ENABLED(CONFIG_IPV6_SIT) && !IS_ENABLED(CONFIG_IPV6_TUNNEL) 159 #define MAX_HEADER LL_MAX_HEADER 160 #else 161 #define MAX_HEADER (LL_MAX_HEADER + 48) 162 #endif 163 164 /* 165 * Old network device statistics. Fields are native words 166 * (unsigned long) so they can be read and written atomically. 167 */ 168 169 struct net_device_stats { 170 unsigned long rx_packets; 171 unsigned long tx_packets; 172 unsigned long rx_bytes; 173 unsigned long tx_bytes; 174 unsigned long rx_errors; 175 unsigned long tx_errors; 176 unsigned long rx_dropped; 177 unsigned long tx_dropped; 178 unsigned long multicast; 179 unsigned long collisions; 180 unsigned long rx_length_errors; 181 unsigned long rx_over_errors; 182 unsigned long rx_crc_errors; 183 unsigned long rx_frame_errors; 184 unsigned long rx_fifo_errors; 185 unsigned long rx_missed_errors; 186 unsigned long tx_aborted_errors; 187 unsigned long tx_carrier_errors; 188 unsigned long tx_fifo_errors; 189 unsigned long tx_heartbeat_errors; 190 unsigned long tx_window_errors; 191 unsigned long rx_compressed; 192 unsigned long tx_compressed; 193 }; 194 195 196 #include <linux/cache.h> 197 #include <linux/skbuff.h> 198 199 #ifdef CONFIG_RPS 200 #include <linux/static_key.h> 201 extern struct static_key_false rps_needed; 202 extern struct static_key_false rfs_needed; 203 #endif 204 205 struct neighbour; 206 struct neigh_parms; 207 struct sk_buff; 208 209 struct netdev_hw_addr { 210 struct list_head list; 211 unsigned char addr[MAX_ADDR_LEN]; 212 unsigned char type; 213 #define NETDEV_HW_ADDR_T_LAN 1 214 #define NETDEV_HW_ADDR_T_SAN 2 215 #define NETDEV_HW_ADDR_T_UNICAST 3 216 #define NETDEV_HW_ADDR_T_MULTICAST 4 217 bool global_use; 218 int sync_cnt; 219 int refcount; 220 int synced; 221 struct rcu_head rcu_head; 222 }; 223 224 struct netdev_hw_addr_list { 225 struct list_head list; 226 int count; 227 }; 228 229 #define netdev_hw_addr_list_count(l) ((l)->count) 230 #define netdev_hw_addr_list_empty(l) (netdev_hw_addr_list_count(l) == 0) 231 #define netdev_hw_addr_list_for_each(ha, l) \ 232 list_for_each_entry(ha, &(l)->list, list) 233 234 #define netdev_uc_count(dev) netdev_hw_addr_list_count(&(dev)->uc) 235 #define netdev_uc_empty(dev) netdev_hw_addr_list_empty(&(dev)->uc) 236 #define netdev_for_each_uc_addr(ha, dev) \ 237 netdev_hw_addr_list_for_each(ha, &(dev)->uc) 238 239 #define netdev_mc_count(dev) netdev_hw_addr_list_count(&(dev)->mc) 240 #define netdev_mc_empty(dev) netdev_hw_addr_list_empty(&(dev)->mc) 241 #define netdev_for_each_mc_addr(ha, dev) \ 242 netdev_hw_addr_list_for_each(ha, &(dev)->mc) 243 244 struct hh_cache { 245 unsigned int hh_len; 246 seqlock_t hh_lock; 247 248 /* cached hardware header; allow for machine alignment needs. */ 249 #define HH_DATA_MOD 16 250 #define HH_DATA_OFF(__len) \ 251 (HH_DATA_MOD - (((__len - 1) & (HH_DATA_MOD - 1)) + 1)) 252 #define HH_DATA_ALIGN(__len) \ 253 (((__len)+(HH_DATA_MOD-1))&~(HH_DATA_MOD - 1)) 254 unsigned long hh_data[HH_DATA_ALIGN(LL_MAX_HEADER) / sizeof(long)]; 255 }; 256 257 /* Reserve HH_DATA_MOD byte-aligned hard_header_len, but at least that much. 258 * Alternative is: 259 * dev->hard_header_len ? (dev->hard_header_len + 260 * (HH_DATA_MOD - 1)) & ~(HH_DATA_MOD - 1) : 0 261 * 262 * We could use other alignment values, but we must maintain the 263 * relationship HH alignment <= LL alignment. 264 */ 265 #define LL_RESERVED_SPACE(dev) \ 266 ((((dev)->hard_header_len+(dev)->needed_headroom)&~(HH_DATA_MOD - 1)) + HH_DATA_MOD) 267 #define LL_RESERVED_SPACE_EXTRA(dev,extra) \ 268 ((((dev)->hard_header_len+(dev)->needed_headroom+(extra))&~(HH_DATA_MOD - 1)) + HH_DATA_MOD) 269 270 struct header_ops { 271 int (*create) (struct sk_buff *skb, struct net_device *dev, 272 unsigned short type, const void *daddr, 273 const void *saddr, unsigned int len); 274 int (*parse)(const struct sk_buff *skb, unsigned char *haddr); 275 int (*cache)(const struct neighbour *neigh, struct hh_cache *hh, __be16 type); 276 void (*cache_update)(struct hh_cache *hh, 277 const struct net_device *dev, 278 const unsigned char *haddr); 279 bool (*validate)(const char *ll_header, unsigned int len); 280 __be16 (*parse_protocol)(const struct sk_buff *skb); 281 }; 282 283 /* These flag bits are private to the generic network queueing 284 * layer; they may not be explicitly referenced by any other 285 * code. 286 */ 287 288 enum netdev_state_t { 289 __LINK_STATE_START, 290 __LINK_STATE_PRESENT, 291 __LINK_STATE_NOCARRIER, 292 __LINK_STATE_LINKWATCH_PENDING, 293 __LINK_STATE_DORMANT, 294 __LINK_STATE_TESTING, 295 }; 296 297 298 /* 299 * This structure holds boot-time configured netdevice settings. They 300 * are then used in the device probing. 301 */ 302 struct netdev_boot_setup { 303 char name[IFNAMSIZ]; 304 struct ifmap map; 305 }; 306 #define NETDEV_BOOT_SETUP_MAX 8 307 308 int __init netdev_boot_setup(char *str); 309 310 struct gro_list { 311 struct list_head list; 312 int count; 313 }; 314 315 /* 316 * size of gro hash buckets, must less than bit number of 317 * napi_struct::gro_bitmask 318 */ 319 #define GRO_HASH_BUCKETS 8 320 321 /* 322 * Structure for NAPI scheduling similar to tasklet but with weighting 323 */ 324 struct napi_struct { 325 /* The poll_list must only be managed by the entity which 326 * changes the state of the NAPI_STATE_SCHED bit. This means 327 * whoever atomically sets that bit can add this napi_struct 328 * to the per-CPU poll_list, and whoever clears that bit 329 * can remove from the list right before clearing the bit. 330 */ 331 struct list_head poll_list; 332 333 unsigned long state; 334 int weight; 335 int defer_hard_irqs_count; 336 unsigned long gro_bitmask; 337 int (*poll)(struct napi_struct *, int); 338 #ifdef CONFIG_NETPOLL 339 int poll_owner; 340 #endif 341 struct net_device *dev; 342 struct gro_list gro_hash[GRO_HASH_BUCKETS]; 343 struct sk_buff *skb; 344 struct list_head rx_list; /* Pending GRO_NORMAL skbs */ 345 int rx_count; /* length of rx_list */ 346 struct hrtimer timer; 347 struct list_head dev_list; 348 struct hlist_node napi_hash_node; 349 unsigned int napi_id; 350 struct task_struct *thread; 351 }; 352 353 enum { 354 NAPI_STATE_SCHED, /* Poll is scheduled */ 355 NAPI_STATE_MISSED, /* reschedule a napi */ 356 NAPI_STATE_DISABLE, /* Disable pending */ 357 NAPI_STATE_NPSVC, /* Netpoll - don't dequeue from poll_list */ 358 NAPI_STATE_LISTED, /* NAPI added to system lists */ 359 NAPI_STATE_NO_BUSY_POLL, /* Do not add in napi_hash, no busy polling */ 360 NAPI_STATE_IN_BUSY_POLL, /* sk_busy_loop() owns this NAPI */ 361 NAPI_STATE_PREFER_BUSY_POLL, /* prefer busy-polling over softirq processing*/ 362 NAPI_STATE_THREADED, /* The poll is performed inside its own thread*/ 363 }; 364 365 enum { 366 NAPIF_STATE_SCHED = BIT(NAPI_STATE_SCHED), 367 NAPIF_STATE_MISSED = BIT(NAPI_STATE_MISSED), 368 NAPIF_STATE_DISABLE = BIT(NAPI_STATE_DISABLE), 369 NAPIF_STATE_NPSVC = BIT(NAPI_STATE_NPSVC), 370 NAPIF_STATE_LISTED = BIT(NAPI_STATE_LISTED), 371 NAPIF_STATE_NO_BUSY_POLL = BIT(NAPI_STATE_NO_BUSY_POLL), 372 NAPIF_STATE_IN_BUSY_POLL = BIT(NAPI_STATE_IN_BUSY_POLL), 373 NAPIF_STATE_PREFER_BUSY_POLL = BIT(NAPI_STATE_PREFER_BUSY_POLL), 374 NAPIF_STATE_THREADED = BIT(NAPI_STATE_THREADED), 375 }; 376 377 enum gro_result { 378 GRO_MERGED, 379 GRO_MERGED_FREE, 380 GRO_HELD, 381 GRO_NORMAL, 382 GRO_CONSUMED, 383 }; 384 typedef enum gro_result gro_result_t; 385 386 /* 387 * enum rx_handler_result - Possible return values for rx_handlers. 388 * @RX_HANDLER_CONSUMED: skb was consumed by rx_handler, do not process it 389 * further. 390 * @RX_HANDLER_ANOTHER: Do another round in receive path. This is indicated in 391 * case skb->dev was changed by rx_handler. 392 * @RX_HANDLER_EXACT: Force exact delivery, no wildcard. 393 * @RX_HANDLER_PASS: Do nothing, pass the skb as if no rx_handler was called. 394 * 395 * rx_handlers are functions called from inside __netif_receive_skb(), to do 396 * special processing of the skb, prior to delivery to protocol handlers. 397 * 398 * Currently, a net_device can only have a single rx_handler registered. Trying 399 * to register a second rx_handler will return -EBUSY. 400 * 401 * To register a rx_handler on a net_device, use netdev_rx_handler_register(). 402 * To unregister a rx_handler on a net_device, use 403 * netdev_rx_handler_unregister(). 404 * 405 * Upon return, rx_handler is expected to tell __netif_receive_skb() what to 406 * do with the skb. 407 * 408 * If the rx_handler consumed the skb in some way, it should return 409 * RX_HANDLER_CONSUMED. This is appropriate when the rx_handler arranged for 410 * the skb to be delivered in some other way. 411 * 412 * If the rx_handler changed skb->dev, to divert the skb to another 413 * net_device, it should return RX_HANDLER_ANOTHER. The rx_handler for the 414 * new device will be called if it exists. 415 * 416 * If the rx_handler decides the skb should be ignored, it should return 417 * RX_HANDLER_EXACT. The skb will only be delivered to protocol handlers that 418 * are registered on exact device (ptype->dev == skb->dev). 419 * 420 * If the rx_handler didn't change skb->dev, but wants the skb to be normally 421 * delivered, it should return RX_HANDLER_PASS. 422 * 423 * A device without a registered rx_handler will behave as if rx_handler 424 * returned RX_HANDLER_PASS. 425 */ 426 427 enum rx_handler_result { 428 RX_HANDLER_CONSUMED, 429 RX_HANDLER_ANOTHER, 430 RX_HANDLER_EXACT, 431 RX_HANDLER_PASS, 432 }; 433 typedef enum rx_handler_result rx_handler_result_t; 434 typedef rx_handler_result_t rx_handler_func_t(struct sk_buff **pskb); 435 436 void __napi_schedule(struct napi_struct *n); 437 void __napi_schedule_irqoff(struct napi_struct *n); 438 439 static inline bool napi_disable_pending(struct napi_struct *n) 440 { 441 return test_bit(NAPI_STATE_DISABLE, &n->state); 442 } 443 444 static inline bool napi_prefer_busy_poll(struct napi_struct *n) 445 { 446 return test_bit(NAPI_STATE_PREFER_BUSY_POLL, &n->state); 447 } 448 449 bool napi_schedule_prep(struct napi_struct *n); 450 451 /** 452 * napi_schedule - schedule NAPI poll 453 * @n: NAPI context 454 * 455 * Schedule NAPI poll routine to be called if it is not already 456 * running. 457 */ 458 static inline void napi_schedule(struct napi_struct *n) 459 { 460 if (napi_schedule_prep(n)) 461 __napi_schedule(n); 462 } 463 464 /** 465 * napi_schedule_irqoff - schedule NAPI poll 466 * @n: NAPI context 467 * 468 * Variant of napi_schedule(), assuming hard irqs are masked. 469 */ 470 static inline void napi_schedule_irqoff(struct napi_struct *n) 471 { 472 if (napi_schedule_prep(n)) 473 __napi_schedule_irqoff(n); 474 } 475 476 /* Try to reschedule poll. Called by dev->poll() after napi_complete(). */ 477 static inline bool napi_reschedule(struct napi_struct *napi) 478 { 479 if (napi_schedule_prep(napi)) { 480 __napi_schedule(napi); 481 return true; 482 } 483 return false; 484 } 485 486 bool napi_complete_done(struct napi_struct *n, int work_done); 487 /** 488 * napi_complete - NAPI processing complete 489 * @n: NAPI context 490 * 491 * Mark NAPI processing as complete. 492 * Consider using napi_complete_done() instead. 493 * Return false if device should avoid rearming interrupts. 494 */ 495 static inline bool napi_complete(struct napi_struct *n) 496 { 497 return napi_complete_done(n, 0); 498 } 499 500 int dev_set_threaded(struct net_device *dev, bool threaded); 501 502 /** 503 * napi_disable - prevent NAPI from scheduling 504 * @n: NAPI context 505 * 506 * Stop NAPI from being scheduled on this context. 507 * Waits till any outstanding processing completes. 508 */ 509 void napi_disable(struct napi_struct *n); 510 511 void napi_enable(struct napi_struct *n); 512 513 /** 514 * napi_synchronize - wait until NAPI is not running 515 * @n: NAPI context 516 * 517 * Wait until NAPI is done being scheduled on this context. 518 * Waits till any outstanding processing completes but 519 * does not disable future activations. 520 */ 521 static inline void napi_synchronize(const struct napi_struct *n) 522 { 523 if (IS_ENABLED(CONFIG_SMP)) 524 while (test_bit(NAPI_STATE_SCHED, &n->state)) 525 msleep(1); 526 else 527 barrier(); 528 } 529 530 /** 531 * napi_if_scheduled_mark_missed - if napi is running, set the 532 * NAPIF_STATE_MISSED 533 * @n: NAPI context 534 * 535 * If napi is running, set the NAPIF_STATE_MISSED, and return true if 536 * NAPI is scheduled. 537 **/ 538 static inline bool napi_if_scheduled_mark_missed(struct napi_struct *n) 539 { 540 unsigned long val, new; 541 542 do { 543 val = READ_ONCE(n->state); 544 if (val & NAPIF_STATE_DISABLE) 545 return true; 546 547 if (!(val & NAPIF_STATE_SCHED)) 548 return false; 549 550 new = val | NAPIF_STATE_MISSED; 551 } while (cmpxchg(&n->state, val, new) != val); 552 553 return true; 554 } 555 556 enum netdev_queue_state_t { 557 __QUEUE_STATE_DRV_XOFF, 558 __QUEUE_STATE_STACK_XOFF, 559 __QUEUE_STATE_FROZEN, 560 }; 561 562 #define QUEUE_STATE_DRV_XOFF (1 << __QUEUE_STATE_DRV_XOFF) 563 #define QUEUE_STATE_STACK_XOFF (1 << __QUEUE_STATE_STACK_XOFF) 564 #define QUEUE_STATE_FROZEN (1 << __QUEUE_STATE_FROZEN) 565 566 #define QUEUE_STATE_ANY_XOFF (QUEUE_STATE_DRV_XOFF | QUEUE_STATE_STACK_XOFF) 567 #define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \ 568 QUEUE_STATE_FROZEN) 569 #define QUEUE_STATE_DRV_XOFF_OR_FROZEN (QUEUE_STATE_DRV_XOFF | \ 570 QUEUE_STATE_FROZEN) 571 572 /* 573 * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue. The 574 * netif_tx_* functions below are used to manipulate this flag. The 575 * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit 576 * queue independently. The netif_xmit_*stopped functions below are called 577 * to check if the queue has been stopped by the driver or stack (either 578 * of the XOFF bits are set in the state). Drivers should not need to call 579 * netif_xmit*stopped functions, they should only be using netif_tx_*. 580 */ 581 582 struct netdev_queue { 583 /* 584 * read-mostly part 585 */ 586 struct net_device *dev; 587 struct Qdisc __rcu *qdisc; 588 struct Qdisc *qdisc_sleeping; 589 #ifdef CONFIG_SYSFS 590 struct kobject kobj; 591 #endif 592 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA) 593 int numa_node; 594 #endif 595 unsigned long tx_maxrate; 596 /* 597 * Number of TX timeouts for this queue 598 * (/sys/class/net/DEV/Q/trans_timeout) 599 */ 600 unsigned long trans_timeout; 601 602 /* Subordinate device that the queue has been assigned to */ 603 struct net_device *sb_dev; 604 #ifdef CONFIG_XDP_SOCKETS 605 struct xsk_buff_pool *pool; 606 #endif 607 /* 608 * write-mostly part 609 */ 610 spinlock_t _xmit_lock ____cacheline_aligned_in_smp; 611 int xmit_lock_owner; 612 /* 613 * Time (in jiffies) of last Tx 614 */ 615 unsigned long trans_start; 616 617 unsigned long state; 618 619 #ifdef CONFIG_BQL 620 struct dql dql; 621 #endif 622 } ____cacheline_aligned_in_smp; 623 624 extern int sysctl_fb_tunnels_only_for_init_net; 625 extern int sysctl_devconf_inherit_init_net; 626 627 /* 628 * sysctl_fb_tunnels_only_for_init_net == 0 : For all netns 629 * == 1 : For initns only 630 * == 2 : For none. 631 */ 632 static inline bool net_has_fallback_tunnels(const struct net *net) 633 { 634 return !IS_ENABLED(CONFIG_SYSCTL) || 635 !sysctl_fb_tunnels_only_for_init_net || 636 (net == &init_net && sysctl_fb_tunnels_only_for_init_net == 1); 637 } 638 639 static inline int netdev_queue_numa_node_read(const struct netdev_queue *q) 640 { 641 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA) 642 return q->numa_node; 643 #else 644 return NUMA_NO_NODE; 645 #endif 646 } 647 648 static inline void netdev_queue_numa_node_write(struct netdev_queue *q, int node) 649 { 650 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA) 651 q->numa_node = node; 652 #endif 653 } 654 655 #ifdef CONFIG_RPS 656 /* 657 * This structure holds an RPS map which can be of variable length. The 658 * map is an array of CPUs. 659 */ 660 struct rps_map { 661 unsigned int len; 662 struct rcu_head rcu; 663 u16 cpus[]; 664 }; 665 #define RPS_MAP_SIZE(_num) (sizeof(struct rps_map) + ((_num) * sizeof(u16))) 666 667 /* 668 * The rps_dev_flow structure contains the mapping of a flow to a CPU, the 669 * tail pointer for that CPU's input queue at the time of last enqueue, and 670 * a hardware filter index. 671 */ 672 struct rps_dev_flow { 673 u16 cpu; 674 u16 filter; 675 unsigned int last_qtail; 676 }; 677 #define RPS_NO_FILTER 0xffff 678 679 /* 680 * The rps_dev_flow_table structure contains a table of flow mappings. 681 */ 682 struct rps_dev_flow_table { 683 unsigned int mask; 684 struct rcu_head rcu; 685 struct rps_dev_flow flows[]; 686 }; 687 #define RPS_DEV_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_dev_flow_table) + \ 688 ((_num) * sizeof(struct rps_dev_flow))) 689 690 /* 691 * The rps_sock_flow_table contains mappings of flows to the last CPU 692 * on which they were processed by the application (set in recvmsg). 693 * Each entry is a 32bit value. Upper part is the high-order bits 694 * of flow hash, lower part is CPU number. 695 * rps_cpu_mask is used to partition the space, depending on number of 696 * possible CPUs : rps_cpu_mask = roundup_pow_of_two(nr_cpu_ids) - 1 697 * For example, if 64 CPUs are possible, rps_cpu_mask = 0x3f, 698 * meaning we use 32-6=26 bits for the hash. 699 */ 700 struct rps_sock_flow_table { 701 u32 mask; 702 703 u32 ents[] ____cacheline_aligned_in_smp; 704 }; 705 #define RPS_SOCK_FLOW_TABLE_SIZE(_num) (offsetof(struct rps_sock_flow_table, ents[_num])) 706 707 #define RPS_NO_CPU 0xffff 708 709 extern u32 rps_cpu_mask; 710 extern struct rps_sock_flow_table __rcu *rps_sock_flow_table; 711 712 static inline void rps_record_sock_flow(struct rps_sock_flow_table *table, 713 u32 hash) 714 { 715 if (table && hash) { 716 unsigned int index = hash & table->mask; 717 u32 val = hash & ~rps_cpu_mask; 718 719 /* We only give a hint, preemption can change CPU under us */ 720 val |= raw_smp_processor_id(); 721 722 if (table->ents[index] != val) 723 table->ents[index] = val; 724 } 725 } 726 727 #ifdef CONFIG_RFS_ACCEL 728 bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index, u32 flow_id, 729 u16 filter_id); 730 #endif 731 #endif /* CONFIG_RPS */ 732 733 /* This structure contains an instance of an RX queue. */ 734 struct netdev_rx_queue { 735 #ifdef CONFIG_RPS 736 struct rps_map __rcu *rps_map; 737 struct rps_dev_flow_table __rcu *rps_flow_table; 738 #endif 739 struct kobject kobj; 740 struct net_device *dev; 741 struct xdp_rxq_info xdp_rxq; 742 #ifdef CONFIG_XDP_SOCKETS 743 struct xsk_buff_pool *pool; 744 #endif 745 } ____cacheline_aligned_in_smp; 746 747 /* 748 * RX queue sysfs structures and functions. 749 */ 750 struct rx_queue_attribute { 751 struct attribute attr; 752 ssize_t (*show)(struct netdev_rx_queue *queue, char *buf); 753 ssize_t (*store)(struct netdev_rx_queue *queue, 754 const char *buf, size_t len); 755 }; 756 757 /* XPS map type and offset of the xps map within net_device->xps_maps[]. */ 758 enum xps_map_type { 759 XPS_CPUS = 0, 760 XPS_RXQS, 761 XPS_MAPS_MAX, 762 }; 763 764 #ifdef CONFIG_XPS 765 /* 766 * This structure holds an XPS map which can be of variable length. The 767 * map is an array of queues. 768 */ 769 struct xps_map { 770 unsigned int len; 771 unsigned int alloc_len; 772 struct rcu_head rcu; 773 u16 queues[]; 774 }; 775 #define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16))) 776 #define XPS_MIN_MAP_ALLOC ((L1_CACHE_ALIGN(offsetof(struct xps_map, queues[1])) \ 777 - sizeof(struct xps_map)) / sizeof(u16)) 778 779 /* 780 * This structure holds all XPS maps for device. Maps are indexed by CPU. 781 * 782 * We keep track of the number of cpus/rxqs used when the struct is allocated, 783 * in nr_ids. This will help not accessing out-of-bound memory. 784 * 785 * We keep track of the number of traffic classes used when the struct is 786 * allocated, in num_tc. This will be used to navigate the maps, to ensure we're 787 * not crossing its upper bound, as the original dev->num_tc can be updated in 788 * the meantime. 789 */ 790 struct xps_dev_maps { 791 struct rcu_head rcu; 792 unsigned int nr_ids; 793 s16 num_tc; 794 struct xps_map __rcu *attr_map[]; /* Either CPUs map or RXQs map */ 795 }; 796 797 #define XPS_CPU_DEV_MAPS_SIZE(_tcs) (sizeof(struct xps_dev_maps) + \ 798 (nr_cpu_ids * (_tcs) * sizeof(struct xps_map *))) 799 800 #define XPS_RXQ_DEV_MAPS_SIZE(_tcs, _rxqs) (sizeof(struct xps_dev_maps) +\ 801 (_rxqs * (_tcs) * sizeof(struct xps_map *))) 802 803 #endif /* CONFIG_XPS */ 804 805 #define TC_MAX_QUEUE 16 806 #define TC_BITMASK 15 807 /* HW offloaded queuing disciplines txq count and offset maps */ 808 struct netdev_tc_txq { 809 u16 count; 810 u16 offset; 811 }; 812 813 #if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE) 814 /* 815 * This structure is to hold information about the device 816 * configured to run FCoE protocol stack. 817 */ 818 struct netdev_fcoe_hbainfo { 819 char manufacturer[64]; 820 char serial_number[64]; 821 char hardware_version[64]; 822 char driver_version[64]; 823 char optionrom_version[64]; 824 char firmware_version[64]; 825 char model[256]; 826 char model_description[256]; 827 }; 828 #endif 829 830 #define MAX_PHYS_ITEM_ID_LEN 32 831 832 /* This structure holds a unique identifier to identify some 833 * physical item (port for example) used by a netdevice. 834 */ 835 struct netdev_phys_item_id { 836 unsigned char id[MAX_PHYS_ITEM_ID_LEN]; 837 unsigned char id_len; 838 }; 839 840 static inline bool netdev_phys_item_id_same(struct netdev_phys_item_id *a, 841 struct netdev_phys_item_id *b) 842 { 843 return a->id_len == b->id_len && 844 memcmp(a->id, b->id, a->id_len) == 0; 845 } 846 847 typedef u16 (*select_queue_fallback_t)(struct net_device *dev, 848 struct sk_buff *skb, 849 struct net_device *sb_dev); 850 851 enum tc_setup_type { 852 TC_SETUP_QDISC_MQPRIO, 853 TC_SETUP_CLSU32, 854 TC_SETUP_CLSFLOWER, 855 TC_SETUP_CLSMATCHALL, 856 TC_SETUP_CLSBPF, 857 TC_SETUP_BLOCK, 858 TC_SETUP_QDISC_CBS, 859 TC_SETUP_QDISC_RED, 860 TC_SETUP_QDISC_PRIO, 861 TC_SETUP_QDISC_MQ, 862 TC_SETUP_QDISC_ETF, 863 TC_SETUP_ROOT_QDISC, 864 TC_SETUP_QDISC_GRED, 865 TC_SETUP_QDISC_TAPRIO, 866 TC_SETUP_FT, 867 TC_SETUP_QDISC_ETS, 868 TC_SETUP_QDISC_TBF, 869 TC_SETUP_QDISC_FIFO, 870 TC_SETUP_QDISC_HTB, 871 }; 872 873 /* These structures hold the attributes of bpf state that are being passed 874 * to the netdevice through the bpf op. 875 */ 876 enum bpf_netdev_command { 877 /* Set or clear a bpf program used in the earliest stages of packet 878 * rx. The prog will have been loaded as BPF_PROG_TYPE_XDP. The callee 879 * is responsible for calling bpf_prog_put on any old progs that are 880 * stored. In case of error, the callee need not release the new prog 881 * reference, but on success it takes ownership and must bpf_prog_put 882 * when it is no longer used. 883 */ 884 XDP_SETUP_PROG, 885 XDP_SETUP_PROG_HW, 886 /* BPF program for offload callbacks, invoked at program load time. */ 887 BPF_OFFLOAD_MAP_ALLOC, 888 BPF_OFFLOAD_MAP_FREE, 889 XDP_SETUP_XSK_POOL, 890 }; 891 892 struct bpf_prog_offload_ops; 893 struct netlink_ext_ack; 894 struct xdp_umem; 895 struct xdp_dev_bulk_queue; 896 struct bpf_xdp_link; 897 898 enum bpf_xdp_mode { 899 XDP_MODE_SKB = 0, 900 XDP_MODE_DRV = 1, 901 XDP_MODE_HW = 2, 902 __MAX_XDP_MODE 903 }; 904 905 struct bpf_xdp_entity { 906 struct bpf_prog *prog; 907 struct bpf_xdp_link *link; 908 }; 909 910 struct netdev_bpf { 911 enum bpf_netdev_command command; 912 union { 913 /* XDP_SETUP_PROG */ 914 struct { 915 u32 flags; 916 struct bpf_prog *prog; 917 struct netlink_ext_ack *extack; 918 }; 919 /* BPF_OFFLOAD_MAP_ALLOC, BPF_OFFLOAD_MAP_FREE */ 920 struct { 921 struct bpf_offloaded_map *offmap; 922 }; 923 /* XDP_SETUP_XSK_POOL */ 924 struct { 925 struct xsk_buff_pool *pool; 926 u16 queue_id; 927 } xsk; 928 }; 929 }; 930 931 /* Flags for ndo_xsk_wakeup. */ 932 #define XDP_WAKEUP_RX (1 << 0) 933 #define XDP_WAKEUP_TX (1 << 1) 934 935 #ifdef CONFIG_XFRM_OFFLOAD 936 struct xfrmdev_ops { 937 int (*xdo_dev_state_add) (struct xfrm_state *x); 938 void (*xdo_dev_state_delete) (struct xfrm_state *x); 939 void (*xdo_dev_state_free) (struct xfrm_state *x); 940 bool (*xdo_dev_offload_ok) (struct sk_buff *skb, 941 struct xfrm_state *x); 942 void (*xdo_dev_state_advance_esn) (struct xfrm_state *x); 943 }; 944 #endif 945 946 struct dev_ifalias { 947 struct rcu_head rcuhead; 948 char ifalias[]; 949 }; 950 951 struct devlink; 952 struct tlsdev_ops; 953 954 struct netdev_name_node { 955 struct hlist_node hlist; 956 struct list_head list; 957 struct net_device *dev; 958 const char *name; 959 }; 960 961 int netdev_name_node_alt_create(struct net_device *dev, const char *name); 962 int netdev_name_node_alt_destroy(struct net_device *dev, const char *name); 963 964 struct netdev_net_notifier { 965 struct list_head list; 966 struct notifier_block *nb; 967 }; 968 969 /* 970 * This structure defines the management hooks for network devices. 971 * The following hooks can be defined; unless noted otherwise, they are 972 * optional and can be filled with a null pointer. 973 * 974 * int (*ndo_init)(struct net_device *dev); 975 * This function is called once when a network device is registered. 976 * The network device can use this for any late stage initialization 977 * or semantic validation. It can fail with an error code which will 978 * be propagated back to register_netdev. 979 * 980 * void (*ndo_uninit)(struct net_device *dev); 981 * This function is called when device is unregistered or when registration 982 * fails. It is not called if init fails. 983 * 984 * int (*ndo_open)(struct net_device *dev); 985 * This function is called when a network device transitions to the up 986 * state. 987 * 988 * int (*ndo_stop)(struct net_device *dev); 989 * This function is called when a network device transitions to the down 990 * state. 991 * 992 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb, 993 * struct net_device *dev); 994 * Called when a packet needs to be transmitted. 995 * Returns NETDEV_TX_OK. Can return NETDEV_TX_BUSY, but you should stop 996 * the queue before that can happen; it's for obsolete devices and weird 997 * corner cases, but the stack really does a non-trivial amount 998 * of useless work if you return NETDEV_TX_BUSY. 999 * Required; cannot be NULL. 1000 * 1001 * netdev_features_t (*ndo_features_check)(struct sk_buff *skb, 1002 * struct net_device *dev 1003 * netdev_features_t features); 1004 * Called by core transmit path to determine if device is capable of 1005 * performing offload operations on a given packet. This is to give 1006 * the device an opportunity to implement any restrictions that cannot 1007 * be otherwise expressed by feature flags. The check is called with 1008 * the set of features that the stack has calculated and it returns 1009 * those the driver believes to be appropriate. 1010 * 1011 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb, 1012 * struct net_device *sb_dev); 1013 * Called to decide which queue to use when device supports multiple 1014 * transmit queues. 1015 * 1016 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags); 1017 * This function is called to allow device receiver to make 1018 * changes to configuration when multicast or promiscuous is enabled. 1019 * 1020 * void (*ndo_set_rx_mode)(struct net_device *dev); 1021 * This function is called device changes address list filtering. 1022 * If driver handles unicast address filtering, it should set 1023 * IFF_UNICAST_FLT in its priv_flags. 1024 * 1025 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr); 1026 * This function is called when the Media Access Control address 1027 * needs to be changed. If this interface is not defined, the 1028 * MAC address can not be changed. 1029 * 1030 * int (*ndo_validate_addr)(struct net_device *dev); 1031 * Test if Media Access Control address is valid for the device. 1032 * 1033 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd); 1034 * Called when a user requests an ioctl which can't be handled by 1035 * the generic interface code. If not defined ioctls return 1036 * not supported error code. 1037 * 1038 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map); 1039 * Used to set network devices bus interface parameters. This interface 1040 * is retained for legacy reasons; new devices should use the bus 1041 * interface (PCI) for low level management. 1042 * 1043 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu); 1044 * Called when a user wants to change the Maximum Transfer Unit 1045 * of a device. 1046 * 1047 * void (*ndo_tx_timeout)(struct net_device *dev, unsigned int txqueue); 1048 * Callback used when the transmitter has not made any progress 1049 * for dev->watchdog ticks. 1050 * 1051 * void (*ndo_get_stats64)(struct net_device *dev, 1052 * struct rtnl_link_stats64 *storage); 1053 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev); 1054 * Called when a user wants to get the network device usage 1055 * statistics. Drivers must do one of the following: 1056 * 1. Define @ndo_get_stats64 to fill in a zero-initialised 1057 * rtnl_link_stats64 structure passed by the caller. 1058 * 2. Define @ndo_get_stats to update a net_device_stats structure 1059 * (which should normally be dev->stats) and return a pointer to 1060 * it. The structure may be changed asynchronously only if each 1061 * field is written atomically. 1062 * 3. Update dev->stats asynchronously and atomically, and define 1063 * neither operation. 1064 * 1065 * bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id) 1066 * Return true if this device supports offload stats of this attr_id. 1067 * 1068 * int (*ndo_get_offload_stats)(int attr_id, const struct net_device *dev, 1069 * void *attr_data) 1070 * Get statistics for offload operations by attr_id. Write it into the 1071 * attr_data pointer. 1072 * 1073 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16 vid); 1074 * If device supports VLAN filtering this function is called when a 1075 * VLAN id is registered. 1076 * 1077 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, __be16 proto, u16 vid); 1078 * If device supports VLAN filtering this function is called when a 1079 * VLAN id is unregistered. 1080 * 1081 * void (*ndo_poll_controller)(struct net_device *dev); 1082 * 1083 * SR-IOV management functions. 1084 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac); 1085 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan, 1086 * u8 qos, __be16 proto); 1087 * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate, 1088 * int max_tx_rate); 1089 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting); 1090 * int (*ndo_set_vf_trust)(struct net_device *dev, int vf, bool setting); 1091 * int (*ndo_get_vf_config)(struct net_device *dev, 1092 * int vf, struct ifla_vf_info *ivf); 1093 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state); 1094 * int (*ndo_set_vf_port)(struct net_device *dev, int vf, 1095 * struct nlattr *port[]); 1096 * 1097 * Enable or disable the VF ability to query its RSS Redirection Table and 1098 * Hash Key. This is needed since on some devices VF share this information 1099 * with PF and querying it may introduce a theoretical security risk. 1100 * int (*ndo_set_vf_rss_query_en)(struct net_device *dev, int vf, bool setting); 1101 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb); 1102 * int (*ndo_setup_tc)(struct net_device *dev, enum tc_setup_type type, 1103 * void *type_data); 1104 * Called to setup any 'tc' scheduler, classifier or action on @dev. 1105 * This is always called from the stack with the rtnl lock held and netif 1106 * tx queues stopped. This allows the netdevice to perform queue 1107 * management safely. 1108 * 1109 * Fiber Channel over Ethernet (FCoE) offload functions. 1110 * int (*ndo_fcoe_enable)(struct net_device *dev); 1111 * Called when the FCoE protocol stack wants to start using LLD for FCoE 1112 * so the underlying device can perform whatever needed configuration or 1113 * initialization to support acceleration of FCoE traffic. 1114 * 1115 * int (*ndo_fcoe_disable)(struct net_device *dev); 1116 * Called when the FCoE protocol stack wants to stop using LLD for FCoE 1117 * so the underlying device can perform whatever needed clean-ups to 1118 * stop supporting acceleration of FCoE traffic. 1119 * 1120 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid, 1121 * struct scatterlist *sgl, unsigned int sgc); 1122 * Called when the FCoE Initiator wants to initialize an I/O that 1123 * is a possible candidate for Direct Data Placement (DDP). The LLD can 1124 * perform necessary setup and returns 1 to indicate the device is set up 1125 * successfully to perform DDP on this I/O, otherwise this returns 0. 1126 * 1127 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid); 1128 * Called when the FCoE Initiator/Target is done with the DDPed I/O as 1129 * indicated by the FC exchange id 'xid', so the underlying device can 1130 * clean up and reuse resources for later DDP requests. 1131 * 1132 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid, 1133 * struct scatterlist *sgl, unsigned int sgc); 1134 * Called when the FCoE Target wants to initialize an I/O that 1135 * is a possible candidate for Direct Data Placement (DDP). The LLD can 1136 * perform necessary setup and returns 1 to indicate the device is set up 1137 * successfully to perform DDP on this I/O, otherwise this returns 0. 1138 * 1139 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev, 1140 * struct netdev_fcoe_hbainfo *hbainfo); 1141 * Called when the FCoE Protocol stack wants information on the underlying 1142 * device. This information is utilized by the FCoE protocol stack to 1143 * register attributes with Fiber Channel management service as per the 1144 * FC-GS Fabric Device Management Information(FDMI) specification. 1145 * 1146 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type); 1147 * Called when the underlying device wants to override default World Wide 1148 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own 1149 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE 1150 * protocol stack to use. 1151 * 1152 * RFS acceleration. 1153 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb, 1154 * u16 rxq_index, u32 flow_id); 1155 * Set hardware filter for RFS. rxq_index is the target queue index; 1156 * flow_id is a flow ID to be passed to rps_may_expire_flow() later. 1157 * Return the filter ID on success, or a negative error code. 1158 * 1159 * Slave management functions (for bridge, bonding, etc). 1160 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev); 1161 * Called to make another netdev an underling. 1162 * 1163 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev); 1164 * Called to release previously enslaved netdev. 1165 * 1166 * struct net_device *(*ndo_get_xmit_slave)(struct net_device *dev, 1167 * struct sk_buff *skb, 1168 * bool all_slaves); 1169 * Get the xmit slave of master device. If all_slaves is true, function 1170 * assume all the slaves can transmit. 1171 * 1172 * Feature/offload setting functions. 1173 * netdev_features_t (*ndo_fix_features)(struct net_device *dev, 1174 * netdev_features_t features); 1175 * Adjusts the requested feature flags according to device-specific 1176 * constraints, and returns the resulting flags. Must not modify 1177 * the device state. 1178 * 1179 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features); 1180 * Called to update device configuration to new features. Passed 1181 * feature set might be less than what was returned by ndo_fix_features()). 1182 * Must return >0 or -errno if it changed dev->features itself. 1183 * 1184 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[], 1185 * struct net_device *dev, 1186 * const unsigned char *addr, u16 vid, u16 flags, 1187 * struct netlink_ext_ack *extack); 1188 * Adds an FDB entry to dev for addr. 1189 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[], 1190 * struct net_device *dev, 1191 * const unsigned char *addr, u16 vid) 1192 * Deletes the FDB entry from dev coresponding to addr. 1193 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb, 1194 * struct net_device *dev, struct net_device *filter_dev, 1195 * int *idx) 1196 * Used to add FDB entries to dump requests. Implementers should add 1197 * entries to skb and update idx with the number of entries. 1198 * 1199 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh, 1200 * u16 flags, struct netlink_ext_ack *extack) 1201 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq, 1202 * struct net_device *dev, u32 filter_mask, 1203 * int nlflags) 1204 * int (*ndo_bridge_dellink)(struct net_device *dev, struct nlmsghdr *nlh, 1205 * u16 flags); 1206 * 1207 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier); 1208 * Called to change device carrier. Soft-devices (like dummy, team, etc) 1209 * which do not represent real hardware may define this to allow their 1210 * userspace components to manage their virtual carrier state. Devices 1211 * that determine carrier state from physical hardware properties (eg 1212 * network cables) or protocol-dependent mechanisms (eg 1213 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function. 1214 * 1215 * int (*ndo_get_phys_port_id)(struct net_device *dev, 1216 * struct netdev_phys_item_id *ppid); 1217 * Called to get ID of physical port of this device. If driver does 1218 * not implement this, it is assumed that the hw is not able to have 1219 * multiple net devices on single physical port. 1220 * 1221 * int (*ndo_get_port_parent_id)(struct net_device *dev, 1222 * struct netdev_phys_item_id *ppid) 1223 * Called to get the parent ID of the physical port of this device. 1224 * 1225 * void* (*ndo_dfwd_add_station)(struct net_device *pdev, 1226 * struct net_device *dev) 1227 * Called by upper layer devices to accelerate switching or other 1228 * station functionality into hardware. 'pdev is the lowerdev 1229 * to use for the offload and 'dev' is the net device that will 1230 * back the offload. Returns a pointer to the private structure 1231 * the upper layer will maintain. 1232 * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv) 1233 * Called by upper layer device to delete the station created 1234 * by 'ndo_dfwd_add_station'. 'pdev' is the net device backing 1235 * the station and priv is the structure returned by the add 1236 * operation. 1237 * int (*ndo_set_tx_maxrate)(struct net_device *dev, 1238 * int queue_index, u32 maxrate); 1239 * Called when a user wants to set a max-rate limitation of specific 1240 * TX queue. 1241 * int (*ndo_get_iflink)(const struct net_device *dev); 1242 * Called to get the iflink value of this device. 1243 * void (*ndo_change_proto_down)(struct net_device *dev, 1244 * bool proto_down); 1245 * This function is used to pass protocol port error state information 1246 * to the switch driver. The switch driver can react to the proto_down 1247 * by doing a phys down on the associated switch port. 1248 * int (*ndo_fill_metadata_dst)(struct net_device *dev, struct sk_buff *skb); 1249 * This function is used to get egress tunnel information for given skb. 1250 * This is useful for retrieving outer tunnel header parameters while 1251 * sampling packet. 1252 * void (*ndo_set_rx_headroom)(struct net_device *dev, int needed_headroom); 1253 * This function is used to specify the headroom that the skb must 1254 * consider when allocation skb during packet reception. Setting 1255 * appropriate rx headroom value allows avoiding skb head copy on 1256 * forward. Setting a negative value resets the rx headroom to the 1257 * default value. 1258 * int (*ndo_bpf)(struct net_device *dev, struct netdev_bpf *bpf); 1259 * This function is used to set or query state related to XDP on the 1260 * netdevice and manage BPF offload. See definition of 1261 * enum bpf_netdev_command for details. 1262 * int (*ndo_xdp_xmit)(struct net_device *dev, int n, struct xdp_frame **xdp, 1263 * u32 flags); 1264 * This function is used to submit @n XDP packets for transmit on a 1265 * netdevice. Returns number of frames successfully transmitted, frames 1266 * that got dropped are freed/returned via xdp_return_frame(). 1267 * Returns negative number, means general error invoking ndo, meaning 1268 * no frames were xmit'ed and core-caller will free all frames. 1269 * int (*ndo_xsk_wakeup)(struct net_device *dev, u32 queue_id, u32 flags); 1270 * This function is used to wake up the softirq, ksoftirqd or kthread 1271 * responsible for sending and/or receiving packets on a specific 1272 * queue id bound to an AF_XDP socket. The flags field specifies if 1273 * only RX, only Tx, or both should be woken up using the flags 1274 * XDP_WAKEUP_RX and XDP_WAKEUP_TX. 1275 * struct devlink_port *(*ndo_get_devlink_port)(struct net_device *dev); 1276 * Get devlink port instance associated with a given netdev. 1277 * Called with a reference on the netdevice and devlink locks only, 1278 * rtnl_lock is not held. 1279 * int (*ndo_tunnel_ctl)(struct net_device *dev, struct ip_tunnel_parm *p, 1280 * int cmd); 1281 * Add, change, delete or get information on an IPv4 tunnel. 1282 * struct net_device *(*ndo_get_peer_dev)(struct net_device *dev); 1283 * If a device is paired with a peer device, return the peer instance. 1284 * The caller must be under RCU read context. 1285 */ 1286 struct net_device_ops { 1287 int (*ndo_init)(struct net_device *dev); 1288 void (*ndo_uninit)(struct net_device *dev); 1289 int (*ndo_open)(struct net_device *dev); 1290 int (*ndo_stop)(struct net_device *dev); 1291 netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb, 1292 struct net_device *dev); 1293 netdev_features_t (*ndo_features_check)(struct sk_buff *skb, 1294 struct net_device *dev, 1295 netdev_features_t features); 1296 u16 (*ndo_select_queue)(struct net_device *dev, 1297 struct sk_buff *skb, 1298 struct net_device *sb_dev); 1299 void (*ndo_change_rx_flags)(struct net_device *dev, 1300 int flags); 1301 void (*ndo_set_rx_mode)(struct net_device *dev); 1302 int (*ndo_set_mac_address)(struct net_device *dev, 1303 void *addr); 1304 int (*ndo_validate_addr)(struct net_device *dev); 1305 int (*ndo_do_ioctl)(struct net_device *dev, 1306 struct ifreq *ifr, int cmd); 1307 int (*ndo_set_config)(struct net_device *dev, 1308 struct ifmap *map); 1309 int (*ndo_change_mtu)(struct net_device *dev, 1310 int new_mtu); 1311 int (*ndo_neigh_setup)(struct net_device *dev, 1312 struct neigh_parms *); 1313 void (*ndo_tx_timeout) (struct net_device *dev, 1314 unsigned int txqueue); 1315 1316 void (*ndo_get_stats64)(struct net_device *dev, 1317 struct rtnl_link_stats64 *storage); 1318 bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id); 1319 int (*ndo_get_offload_stats)(int attr_id, 1320 const struct net_device *dev, 1321 void *attr_data); 1322 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev); 1323 1324 int (*ndo_vlan_rx_add_vid)(struct net_device *dev, 1325 __be16 proto, u16 vid); 1326 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, 1327 __be16 proto, u16 vid); 1328 #ifdef CONFIG_NET_POLL_CONTROLLER 1329 void (*ndo_poll_controller)(struct net_device *dev); 1330 int (*ndo_netpoll_setup)(struct net_device *dev, 1331 struct netpoll_info *info); 1332 void (*ndo_netpoll_cleanup)(struct net_device *dev); 1333 #endif 1334 int (*ndo_set_vf_mac)(struct net_device *dev, 1335 int queue, u8 *mac); 1336 int (*ndo_set_vf_vlan)(struct net_device *dev, 1337 int queue, u16 vlan, 1338 u8 qos, __be16 proto); 1339 int (*ndo_set_vf_rate)(struct net_device *dev, 1340 int vf, int min_tx_rate, 1341 int max_tx_rate); 1342 int (*ndo_set_vf_spoofchk)(struct net_device *dev, 1343 int vf, bool setting); 1344 int (*ndo_set_vf_trust)(struct net_device *dev, 1345 int vf, bool setting); 1346 int (*ndo_get_vf_config)(struct net_device *dev, 1347 int vf, 1348 struct ifla_vf_info *ivf); 1349 int (*ndo_set_vf_link_state)(struct net_device *dev, 1350 int vf, int link_state); 1351 int (*ndo_get_vf_stats)(struct net_device *dev, 1352 int vf, 1353 struct ifla_vf_stats 1354 *vf_stats); 1355 int (*ndo_set_vf_port)(struct net_device *dev, 1356 int vf, 1357 struct nlattr *port[]); 1358 int (*ndo_get_vf_port)(struct net_device *dev, 1359 int vf, struct sk_buff *skb); 1360 int (*ndo_get_vf_guid)(struct net_device *dev, 1361 int vf, 1362 struct ifla_vf_guid *node_guid, 1363 struct ifla_vf_guid *port_guid); 1364 int (*ndo_set_vf_guid)(struct net_device *dev, 1365 int vf, u64 guid, 1366 int guid_type); 1367 int (*ndo_set_vf_rss_query_en)( 1368 struct net_device *dev, 1369 int vf, bool setting); 1370 int (*ndo_setup_tc)(struct net_device *dev, 1371 enum tc_setup_type type, 1372 void *type_data); 1373 #if IS_ENABLED(CONFIG_FCOE) 1374 int (*ndo_fcoe_enable)(struct net_device *dev); 1375 int (*ndo_fcoe_disable)(struct net_device *dev); 1376 int (*ndo_fcoe_ddp_setup)(struct net_device *dev, 1377 u16 xid, 1378 struct scatterlist *sgl, 1379 unsigned int sgc); 1380 int (*ndo_fcoe_ddp_done)(struct net_device *dev, 1381 u16 xid); 1382 int (*ndo_fcoe_ddp_target)(struct net_device *dev, 1383 u16 xid, 1384 struct scatterlist *sgl, 1385 unsigned int sgc); 1386 int (*ndo_fcoe_get_hbainfo)(struct net_device *dev, 1387 struct netdev_fcoe_hbainfo *hbainfo); 1388 #endif 1389 1390 #if IS_ENABLED(CONFIG_LIBFCOE) 1391 #define NETDEV_FCOE_WWNN 0 1392 #define NETDEV_FCOE_WWPN 1 1393 int (*ndo_fcoe_get_wwn)(struct net_device *dev, 1394 u64 *wwn, int type); 1395 #endif 1396 1397 #ifdef CONFIG_RFS_ACCEL 1398 int (*ndo_rx_flow_steer)(struct net_device *dev, 1399 const struct sk_buff *skb, 1400 u16 rxq_index, 1401 u32 flow_id); 1402 #endif 1403 int (*ndo_add_slave)(struct net_device *dev, 1404 struct net_device *slave_dev, 1405 struct netlink_ext_ack *extack); 1406 int (*ndo_del_slave)(struct net_device *dev, 1407 struct net_device *slave_dev); 1408 struct net_device* (*ndo_get_xmit_slave)(struct net_device *dev, 1409 struct sk_buff *skb, 1410 bool all_slaves); 1411 struct net_device* (*ndo_sk_get_lower_dev)(struct net_device *dev, 1412 struct sock *sk); 1413 netdev_features_t (*ndo_fix_features)(struct net_device *dev, 1414 netdev_features_t features); 1415 int (*ndo_set_features)(struct net_device *dev, 1416 netdev_features_t features); 1417 int (*ndo_neigh_construct)(struct net_device *dev, 1418 struct neighbour *n); 1419 void (*ndo_neigh_destroy)(struct net_device *dev, 1420 struct neighbour *n); 1421 1422 int (*ndo_fdb_add)(struct ndmsg *ndm, 1423 struct nlattr *tb[], 1424 struct net_device *dev, 1425 const unsigned char *addr, 1426 u16 vid, 1427 u16 flags, 1428 struct netlink_ext_ack *extack); 1429 int (*ndo_fdb_del)(struct ndmsg *ndm, 1430 struct nlattr *tb[], 1431 struct net_device *dev, 1432 const unsigned char *addr, 1433 u16 vid); 1434 int (*ndo_fdb_dump)(struct sk_buff *skb, 1435 struct netlink_callback *cb, 1436 struct net_device *dev, 1437 struct net_device *filter_dev, 1438 int *idx); 1439 int (*ndo_fdb_get)(struct sk_buff *skb, 1440 struct nlattr *tb[], 1441 struct net_device *dev, 1442 const unsigned char *addr, 1443 u16 vid, u32 portid, u32 seq, 1444 struct netlink_ext_ack *extack); 1445 int (*ndo_bridge_setlink)(struct net_device *dev, 1446 struct nlmsghdr *nlh, 1447 u16 flags, 1448 struct netlink_ext_ack *extack); 1449 int (*ndo_bridge_getlink)(struct sk_buff *skb, 1450 u32 pid, u32 seq, 1451 struct net_device *dev, 1452 u32 filter_mask, 1453 int nlflags); 1454 int (*ndo_bridge_dellink)(struct net_device *dev, 1455 struct nlmsghdr *nlh, 1456 u16 flags); 1457 int (*ndo_change_carrier)(struct net_device *dev, 1458 bool new_carrier); 1459 int (*ndo_get_phys_port_id)(struct net_device *dev, 1460 struct netdev_phys_item_id *ppid); 1461 int (*ndo_get_port_parent_id)(struct net_device *dev, 1462 struct netdev_phys_item_id *ppid); 1463 int (*ndo_get_phys_port_name)(struct net_device *dev, 1464 char *name, size_t len); 1465 void* (*ndo_dfwd_add_station)(struct net_device *pdev, 1466 struct net_device *dev); 1467 void (*ndo_dfwd_del_station)(struct net_device *pdev, 1468 void *priv); 1469 1470 int (*ndo_set_tx_maxrate)(struct net_device *dev, 1471 int queue_index, 1472 u32 maxrate); 1473 int (*ndo_get_iflink)(const struct net_device *dev); 1474 int (*ndo_change_proto_down)(struct net_device *dev, 1475 bool proto_down); 1476 int (*ndo_fill_metadata_dst)(struct net_device *dev, 1477 struct sk_buff *skb); 1478 void (*ndo_set_rx_headroom)(struct net_device *dev, 1479 int needed_headroom); 1480 int (*ndo_bpf)(struct net_device *dev, 1481 struct netdev_bpf *bpf); 1482 int (*ndo_xdp_xmit)(struct net_device *dev, int n, 1483 struct xdp_frame **xdp, 1484 u32 flags); 1485 int (*ndo_xsk_wakeup)(struct net_device *dev, 1486 u32 queue_id, u32 flags); 1487 struct devlink_port * (*ndo_get_devlink_port)(struct net_device *dev); 1488 int (*ndo_tunnel_ctl)(struct net_device *dev, 1489 struct ip_tunnel_parm *p, int cmd); 1490 struct net_device * (*ndo_get_peer_dev)(struct net_device *dev); 1491 }; 1492 1493 /** 1494 * enum netdev_priv_flags - &struct net_device priv_flags 1495 * 1496 * These are the &struct net_device, they are only set internally 1497 * by drivers and used in the kernel. These flags are invisible to 1498 * userspace; this means that the order of these flags can change 1499 * during any kernel release. 1500 * 1501 * You should have a pretty good reason to be extending these flags. 1502 * 1503 * @IFF_802_1Q_VLAN: 802.1Q VLAN device 1504 * @IFF_EBRIDGE: Ethernet bridging device 1505 * @IFF_BONDING: bonding master or slave 1506 * @IFF_ISATAP: ISATAP interface (RFC4214) 1507 * @IFF_WAN_HDLC: WAN HDLC device 1508 * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to 1509 * release skb->dst 1510 * @IFF_DONT_BRIDGE: disallow bridging this ether dev 1511 * @IFF_DISABLE_NETPOLL: disable netpoll at run-time 1512 * @IFF_MACVLAN_PORT: device used as macvlan port 1513 * @IFF_BRIDGE_PORT: device used as bridge port 1514 * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port 1515 * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit 1516 * @IFF_UNICAST_FLT: Supports unicast filtering 1517 * @IFF_TEAM_PORT: device used as team port 1518 * @IFF_SUPP_NOFCS: device supports sending custom FCS 1519 * @IFF_LIVE_ADDR_CHANGE: device supports hardware address 1520 * change when it's running 1521 * @IFF_MACVLAN: Macvlan device 1522 * @IFF_XMIT_DST_RELEASE_PERM: IFF_XMIT_DST_RELEASE not taking into account 1523 * underlying stacked devices 1524 * @IFF_L3MDEV_MASTER: device is an L3 master device 1525 * @IFF_NO_QUEUE: device can run without qdisc attached 1526 * @IFF_OPENVSWITCH: device is a Open vSwitch master 1527 * @IFF_L3MDEV_SLAVE: device is enslaved to an L3 master device 1528 * @IFF_TEAM: device is a team device 1529 * @IFF_RXFH_CONFIGURED: device has had Rx Flow indirection table configured 1530 * @IFF_PHONY_HEADROOM: the headroom value is controlled by an external 1531 * entity (i.e. the master device for bridged veth) 1532 * @IFF_MACSEC: device is a MACsec device 1533 * @IFF_NO_RX_HANDLER: device doesn't support the rx_handler hook 1534 * @IFF_FAILOVER: device is a failover master device 1535 * @IFF_FAILOVER_SLAVE: device is lower dev of a failover master device 1536 * @IFF_L3MDEV_RX_HANDLER: only invoke the rx handler of L3 master device 1537 * @IFF_LIVE_RENAME_OK: rename is allowed while device is up and running 1538 * @IFF_TX_SKB_NO_LINEAR: device/driver is capable of xmitting frames with 1539 * skb_headlen(skb) == 0 (data starts from frag0) 1540 */ 1541 enum netdev_priv_flags { 1542 IFF_802_1Q_VLAN = 1<<0, 1543 IFF_EBRIDGE = 1<<1, 1544 IFF_BONDING = 1<<2, 1545 IFF_ISATAP = 1<<3, 1546 IFF_WAN_HDLC = 1<<4, 1547 IFF_XMIT_DST_RELEASE = 1<<5, 1548 IFF_DONT_BRIDGE = 1<<6, 1549 IFF_DISABLE_NETPOLL = 1<<7, 1550 IFF_MACVLAN_PORT = 1<<8, 1551 IFF_BRIDGE_PORT = 1<<9, 1552 IFF_OVS_DATAPATH = 1<<10, 1553 IFF_TX_SKB_SHARING = 1<<11, 1554 IFF_UNICAST_FLT = 1<<12, 1555 IFF_TEAM_PORT = 1<<13, 1556 IFF_SUPP_NOFCS = 1<<14, 1557 IFF_LIVE_ADDR_CHANGE = 1<<15, 1558 IFF_MACVLAN = 1<<16, 1559 IFF_XMIT_DST_RELEASE_PERM = 1<<17, 1560 IFF_L3MDEV_MASTER = 1<<18, 1561 IFF_NO_QUEUE = 1<<19, 1562 IFF_OPENVSWITCH = 1<<20, 1563 IFF_L3MDEV_SLAVE = 1<<21, 1564 IFF_TEAM = 1<<22, 1565 IFF_RXFH_CONFIGURED = 1<<23, 1566 IFF_PHONY_HEADROOM = 1<<24, 1567 IFF_MACSEC = 1<<25, 1568 IFF_NO_RX_HANDLER = 1<<26, 1569 IFF_FAILOVER = 1<<27, 1570 IFF_FAILOVER_SLAVE = 1<<28, 1571 IFF_L3MDEV_RX_HANDLER = 1<<29, 1572 IFF_LIVE_RENAME_OK = 1<<30, 1573 IFF_TX_SKB_NO_LINEAR = 1<<31, 1574 }; 1575 1576 #define IFF_802_1Q_VLAN IFF_802_1Q_VLAN 1577 #define IFF_EBRIDGE IFF_EBRIDGE 1578 #define IFF_BONDING IFF_BONDING 1579 #define IFF_ISATAP IFF_ISATAP 1580 #define IFF_WAN_HDLC IFF_WAN_HDLC 1581 #define IFF_XMIT_DST_RELEASE IFF_XMIT_DST_RELEASE 1582 #define IFF_DONT_BRIDGE IFF_DONT_BRIDGE 1583 #define IFF_DISABLE_NETPOLL IFF_DISABLE_NETPOLL 1584 #define IFF_MACVLAN_PORT IFF_MACVLAN_PORT 1585 #define IFF_BRIDGE_PORT IFF_BRIDGE_PORT 1586 #define IFF_OVS_DATAPATH IFF_OVS_DATAPATH 1587 #define IFF_TX_SKB_SHARING IFF_TX_SKB_SHARING 1588 #define IFF_UNICAST_FLT IFF_UNICAST_FLT 1589 #define IFF_TEAM_PORT IFF_TEAM_PORT 1590 #define IFF_SUPP_NOFCS IFF_SUPP_NOFCS 1591 #define IFF_LIVE_ADDR_CHANGE IFF_LIVE_ADDR_CHANGE 1592 #define IFF_MACVLAN IFF_MACVLAN 1593 #define IFF_XMIT_DST_RELEASE_PERM IFF_XMIT_DST_RELEASE_PERM 1594 #define IFF_L3MDEV_MASTER IFF_L3MDEV_MASTER 1595 #define IFF_NO_QUEUE IFF_NO_QUEUE 1596 #define IFF_OPENVSWITCH IFF_OPENVSWITCH 1597 #define IFF_L3MDEV_SLAVE IFF_L3MDEV_SLAVE 1598 #define IFF_TEAM IFF_TEAM 1599 #define IFF_RXFH_CONFIGURED IFF_RXFH_CONFIGURED 1600 #define IFF_PHONY_HEADROOM IFF_PHONY_HEADROOM 1601 #define IFF_MACSEC IFF_MACSEC 1602 #define IFF_NO_RX_HANDLER IFF_NO_RX_HANDLER 1603 #define IFF_FAILOVER IFF_FAILOVER 1604 #define IFF_FAILOVER_SLAVE IFF_FAILOVER_SLAVE 1605 #define IFF_L3MDEV_RX_HANDLER IFF_L3MDEV_RX_HANDLER 1606 #define IFF_LIVE_RENAME_OK IFF_LIVE_RENAME_OK 1607 #define IFF_TX_SKB_NO_LINEAR IFF_TX_SKB_NO_LINEAR 1608 1609 /* Specifies the type of the struct net_device::ml_priv pointer */ 1610 enum netdev_ml_priv_type { 1611 ML_PRIV_NONE, 1612 ML_PRIV_CAN, 1613 }; 1614 1615 /** 1616 * struct net_device - The DEVICE structure. 1617 * 1618 * Actually, this whole structure is a big mistake. It mixes I/O 1619 * data with strictly "high-level" data, and it has to know about 1620 * almost every data structure used in the INET module. 1621 * 1622 * @name: This is the first field of the "visible" part of this structure 1623 * (i.e. as seen by users in the "Space.c" file). It is the name 1624 * of the interface. 1625 * 1626 * @name_node: Name hashlist node 1627 * @ifalias: SNMP alias 1628 * @mem_end: Shared memory end 1629 * @mem_start: Shared memory start 1630 * @base_addr: Device I/O address 1631 * @irq: Device IRQ number 1632 * 1633 * @state: Generic network queuing layer state, see netdev_state_t 1634 * @dev_list: The global list of network devices 1635 * @napi_list: List entry used for polling NAPI devices 1636 * @unreg_list: List entry when we are unregistering the 1637 * device; see the function unregister_netdev 1638 * @close_list: List entry used when we are closing the device 1639 * @ptype_all: Device-specific packet handlers for all protocols 1640 * @ptype_specific: Device-specific, protocol-specific packet handlers 1641 * 1642 * @adj_list: Directly linked devices, like slaves for bonding 1643 * @features: Currently active device features 1644 * @hw_features: User-changeable features 1645 * 1646 * @wanted_features: User-requested features 1647 * @vlan_features: Mask of features inheritable by VLAN devices 1648 * 1649 * @hw_enc_features: Mask of features inherited by encapsulating devices 1650 * This field indicates what encapsulation 1651 * offloads the hardware is capable of doing, 1652 * and drivers will need to set them appropriately. 1653 * 1654 * @mpls_features: Mask of features inheritable by MPLS 1655 * @gso_partial_features: value(s) from NETIF_F_GSO\* 1656 * 1657 * @ifindex: interface index 1658 * @group: The group the device belongs to 1659 * 1660 * @stats: Statistics struct, which was left as a legacy, use 1661 * rtnl_link_stats64 instead 1662 * 1663 * @rx_dropped: Dropped packets by core network, 1664 * do not use this in drivers 1665 * @tx_dropped: Dropped packets by core network, 1666 * do not use this in drivers 1667 * @rx_nohandler: nohandler dropped packets by core network on 1668 * inactive devices, do not use this in drivers 1669 * @carrier_up_count: Number of times the carrier has been up 1670 * @carrier_down_count: Number of times the carrier has been down 1671 * 1672 * @wireless_handlers: List of functions to handle Wireless Extensions, 1673 * instead of ioctl, 1674 * see <net/iw_handler.h> for details. 1675 * @wireless_data: Instance data managed by the core of wireless extensions 1676 * 1677 * @netdev_ops: Includes several pointers to callbacks, 1678 * if one wants to override the ndo_*() functions 1679 * @ethtool_ops: Management operations 1680 * @l3mdev_ops: Layer 3 master device operations 1681 * @ndisc_ops: Includes callbacks for different IPv6 neighbour 1682 * discovery handling. Necessary for e.g. 6LoWPAN. 1683 * @xfrmdev_ops: Transformation offload operations 1684 * @tlsdev_ops: Transport Layer Security offload operations 1685 * @header_ops: Includes callbacks for creating,parsing,caching,etc 1686 * of Layer 2 headers. 1687 * 1688 * @flags: Interface flags (a la BSD) 1689 * @priv_flags: Like 'flags' but invisible to userspace, 1690 * see if.h for the definitions 1691 * @gflags: Global flags ( kept as legacy ) 1692 * @padded: How much padding added by alloc_netdev() 1693 * @operstate: RFC2863 operstate 1694 * @link_mode: Mapping policy to operstate 1695 * @if_port: Selectable AUI, TP, ... 1696 * @dma: DMA channel 1697 * @mtu: Interface MTU value 1698 * @min_mtu: Interface Minimum MTU value 1699 * @max_mtu: Interface Maximum MTU value 1700 * @type: Interface hardware type 1701 * @hard_header_len: Maximum hardware header length. 1702 * @min_header_len: Minimum hardware header length 1703 * 1704 * @needed_headroom: Extra headroom the hardware may need, but not in all 1705 * cases can this be guaranteed 1706 * @needed_tailroom: Extra tailroom the hardware may need, but not in all 1707 * cases can this be guaranteed. Some cases also use 1708 * LL_MAX_HEADER instead to allocate the skb 1709 * 1710 * interface address info: 1711 * 1712 * @perm_addr: Permanent hw address 1713 * @addr_assign_type: Hw address assignment type 1714 * @addr_len: Hardware address length 1715 * @upper_level: Maximum depth level of upper devices. 1716 * @lower_level: Maximum depth level of lower devices. 1717 * @neigh_priv_len: Used in neigh_alloc() 1718 * @dev_id: Used to differentiate devices that share 1719 * the same link layer address 1720 * @dev_port: Used to differentiate devices that share 1721 * the same function 1722 * @addr_list_lock: XXX: need comments on this one 1723 * @name_assign_type: network interface name assignment type 1724 * @uc_promisc: Counter that indicates promiscuous mode 1725 * has been enabled due to the need to listen to 1726 * additional unicast addresses in a device that 1727 * does not implement ndo_set_rx_mode() 1728 * @uc: unicast mac addresses 1729 * @mc: multicast mac addresses 1730 * @dev_addrs: list of device hw addresses 1731 * @queues_kset: Group of all Kobjects in the Tx and RX queues 1732 * @promiscuity: Number of times the NIC is told to work in 1733 * promiscuous mode; if it becomes 0 the NIC will 1734 * exit promiscuous mode 1735 * @allmulti: Counter, enables or disables allmulticast mode 1736 * 1737 * @vlan_info: VLAN info 1738 * @dsa_ptr: dsa specific data 1739 * @tipc_ptr: TIPC specific data 1740 * @atalk_ptr: AppleTalk link 1741 * @ip_ptr: IPv4 specific data 1742 * @dn_ptr: DECnet specific data 1743 * @ip6_ptr: IPv6 specific data 1744 * @ax25_ptr: AX.25 specific data 1745 * @ieee80211_ptr: IEEE 802.11 specific data, assign before registering 1746 * @ieee802154_ptr: IEEE 802.15.4 low-rate Wireless Personal Area Network 1747 * device struct 1748 * @mpls_ptr: mpls_dev struct pointer 1749 * 1750 * @dev_addr: Hw address (before bcast, 1751 * because most packets are unicast) 1752 * 1753 * @_rx: Array of RX queues 1754 * @num_rx_queues: Number of RX queues 1755 * allocated at register_netdev() time 1756 * @real_num_rx_queues: Number of RX queues currently active in device 1757 * @xdp_prog: XDP sockets filter program pointer 1758 * @gro_flush_timeout: timeout for GRO layer in NAPI 1759 * @napi_defer_hard_irqs: If not zero, provides a counter that would 1760 * allow to avoid NIC hard IRQ, on busy queues. 1761 * 1762 * @rx_handler: handler for received packets 1763 * @rx_handler_data: XXX: need comments on this one 1764 * @miniq_ingress: ingress/clsact qdisc specific data for 1765 * ingress processing 1766 * @ingress_queue: XXX: need comments on this one 1767 * @nf_hooks_ingress: netfilter hooks executed for ingress packets 1768 * @broadcast: hw bcast address 1769 * 1770 * @rx_cpu_rmap: CPU reverse-mapping for RX completion interrupts, 1771 * indexed by RX queue number. Assigned by driver. 1772 * This must only be set if the ndo_rx_flow_steer 1773 * operation is defined 1774 * @index_hlist: Device index hash chain 1775 * 1776 * @_tx: Array of TX queues 1777 * @num_tx_queues: Number of TX queues allocated at alloc_netdev_mq() time 1778 * @real_num_tx_queues: Number of TX queues currently active in device 1779 * @qdisc: Root qdisc from userspace point of view 1780 * @tx_queue_len: Max frames per queue allowed 1781 * @tx_global_lock: XXX: need comments on this one 1782 * @xdp_bulkq: XDP device bulk queue 1783 * @xps_maps: all CPUs/RXQs maps for XPS device 1784 * 1785 * @xps_maps: XXX: need comments on this one 1786 * @miniq_egress: clsact qdisc specific data for 1787 * egress processing 1788 * @qdisc_hash: qdisc hash table 1789 * @watchdog_timeo: Represents the timeout that is used by 1790 * the watchdog (see dev_watchdog()) 1791 * @watchdog_timer: List of timers 1792 * 1793 * @proto_down_reason: reason a netdev interface is held down 1794 * @pcpu_refcnt: Number of references to this device 1795 * @dev_refcnt: Number of references to this device 1796 * @todo_list: Delayed register/unregister 1797 * @link_watch_list: XXX: need comments on this one 1798 * 1799 * @reg_state: Register/unregister state machine 1800 * @dismantle: Device is going to be freed 1801 * @rtnl_link_state: This enum represents the phases of creating 1802 * a new link 1803 * 1804 * @needs_free_netdev: Should unregister perform free_netdev? 1805 * @priv_destructor: Called from unregister 1806 * @npinfo: XXX: need comments on this one 1807 * @nd_net: Network namespace this network device is inside 1808 * 1809 * @ml_priv: Mid-layer private 1810 * @ml_priv_type: Mid-layer private type 1811 * @lstats: Loopback statistics 1812 * @tstats: Tunnel statistics 1813 * @dstats: Dummy statistics 1814 * @vstats: Virtual ethernet statistics 1815 * 1816 * @garp_port: GARP 1817 * @mrp_port: MRP 1818 * 1819 * @dev: Class/net/name entry 1820 * @sysfs_groups: Space for optional device, statistics and wireless 1821 * sysfs groups 1822 * 1823 * @sysfs_rx_queue_group: Space for optional per-rx queue attributes 1824 * @rtnl_link_ops: Rtnl_link_ops 1825 * 1826 * @gso_max_size: Maximum size of generic segmentation offload 1827 * @gso_max_segs: Maximum number of segments that can be passed to the 1828 * NIC for GSO 1829 * 1830 * @dcbnl_ops: Data Center Bridging netlink ops 1831 * @num_tc: Number of traffic classes in the net device 1832 * @tc_to_txq: XXX: need comments on this one 1833 * @prio_tc_map: XXX: need comments on this one 1834 * 1835 * @fcoe_ddp_xid: Max exchange id for FCoE LRO by ddp 1836 * 1837 * @priomap: XXX: need comments on this one 1838 * @phydev: Physical device may attach itself 1839 * for hardware timestamping 1840 * @sfp_bus: attached &struct sfp_bus structure. 1841 * 1842 * @qdisc_tx_busylock: lockdep class annotating Qdisc->busylock spinlock 1843 * @qdisc_running_key: lockdep class annotating Qdisc->running seqcount 1844 * 1845 * @proto_down: protocol port state information can be sent to the 1846 * switch driver and used to set the phys state of the 1847 * switch port. 1848 * 1849 * @wol_enabled: Wake-on-LAN is enabled 1850 * 1851 * @threaded: napi threaded mode is enabled 1852 * 1853 * @net_notifier_list: List of per-net netdev notifier block 1854 * that follow this device when it is moved 1855 * to another network namespace. 1856 * 1857 * @macsec_ops: MACsec offloading ops 1858 * 1859 * @udp_tunnel_nic_info: static structure describing the UDP tunnel 1860 * offload capabilities of the device 1861 * @udp_tunnel_nic: UDP tunnel offload state 1862 * @xdp_state: stores info on attached XDP BPF programs 1863 * 1864 * @nested_level: Used as as a parameter of spin_lock_nested() of 1865 * dev->addr_list_lock. 1866 * @unlink_list: As netif_addr_lock() can be called recursively, 1867 * keep a list of interfaces to be deleted. 1868 * 1869 * FIXME: cleanup struct net_device such that network protocol info 1870 * moves out. 1871 */ 1872 1873 struct net_device { 1874 char name[IFNAMSIZ]; 1875 struct netdev_name_node *name_node; 1876 struct dev_ifalias __rcu *ifalias; 1877 /* 1878 * I/O specific fields 1879 * FIXME: Merge these and struct ifmap into one 1880 */ 1881 unsigned long mem_end; 1882 unsigned long mem_start; 1883 unsigned long base_addr; 1884 1885 /* 1886 * Some hardware also needs these fields (state,dev_list, 1887 * napi_list,unreg_list,close_list) but they are not 1888 * part of the usual set specified in Space.c. 1889 */ 1890 1891 unsigned long state; 1892 1893 struct list_head dev_list; 1894 struct list_head napi_list; 1895 struct list_head unreg_list; 1896 struct list_head close_list; 1897 struct list_head ptype_all; 1898 struct list_head ptype_specific; 1899 1900 struct { 1901 struct list_head upper; 1902 struct list_head lower; 1903 } adj_list; 1904 1905 /* Read-mostly cache-line for fast-path access */ 1906 unsigned int flags; 1907 unsigned int priv_flags; 1908 const struct net_device_ops *netdev_ops; 1909 int ifindex; 1910 unsigned short gflags; 1911 unsigned short hard_header_len; 1912 1913 /* Note : dev->mtu is often read without holding a lock. 1914 * Writers usually hold RTNL. 1915 * It is recommended to use READ_ONCE() to annotate the reads, 1916 * and to use WRITE_ONCE() to annotate the writes. 1917 */ 1918 unsigned int mtu; 1919 unsigned short needed_headroom; 1920 unsigned short needed_tailroom; 1921 1922 netdev_features_t features; 1923 netdev_features_t hw_features; 1924 netdev_features_t wanted_features; 1925 netdev_features_t vlan_features; 1926 netdev_features_t hw_enc_features; 1927 netdev_features_t mpls_features; 1928 netdev_features_t gso_partial_features; 1929 1930 unsigned int min_mtu; 1931 unsigned int max_mtu; 1932 unsigned short type; 1933 unsigned char min_header_len; 1934 unsigned char name_assign_type; 1935 1936 int group; 1937 1938 struct net_device_stats stats; /* not used by modern drivers */ 1939 1940 atomic_long_t rx_dropped; 1941 atomic_long_t tx_dropped; 1942 atomic_long_t rx_nohandler; 1943 1944 /* Stats to monitor link on/off, flapping */ 1945 atomic_t carrier_up_count; 1946 atomic_t carrier_down_count; 1947 1948 #ifdef CONFIG_WIRELESS_EXT 1949 const struct iw_handler_def *wireless_handlers; 1950 struct iw_public_data *wireless_data; 1951 #endif 1952 const struct ethtool_ops *ethtool_ops; 1953 #ifdef CONFIG_NET_L3_MASTER_DEV 1954 const struct l3mdev_ops *l3mdev_ops; 1955 #endif 1956 #if IS_ENABLED(CONFIG_IPV6) 1957 const struct ndisc_ops *ndisc_ops; 1958 #endif 1959 1960 #ifdef CONFIG_XFRM_OFFLOAD 1961 const struct xfrmdev_ops *xfrmdev_ops; 1962 #endif 1963 1964 #if IS_ENABLED(CONFIG_TLS_DEVICE) 1965 const struct tlsdev_ops *tlsdev_ops; 1966 #endif 1967 1968 const struct header_ops *header_ops; 1969 1970 unsigned char operstate; 1971 unsigned char link_mode; 1972 1973 unsigned char if_port; 1974 unsigned char dma; 1975 1976 /* Interface address info. */ 1977 unsigned char perm_addr[MAX_ADDR_LEN]; 1978 unsigned char addr_assign_type; 1979 unsigned char addr_len; 1980 unsigned char upper_level; 1981 unsigned char lower_level; 1982 1983 unsigned short neigh_priv_len; 1984 unsigned short dev_id; 1985 unsigned short dev_port; 1986 unsigned short padded; 1987 1988 spinlock_t addr_list_lock; 1989 int irq; 1990 1991 struct netdev_hw_addr_list uc; 1992 struct netdev_hw_addr_list mc; 1993 struct netdev_hw_addr_list dev_addrs; 1994 1995 #ifdef CONFIG_SYSFS 1996 struct kset *queues_kset; 1997 #endif 1998 #ifdef CONFIG_LOCKDEP 1999 struct list_head unlink_list; 2000 #endif 2001 unsigned int promiscuity; 2002 unsigned int allmulti; 2003 bool uc_promisc; 2004 #ifdef CONFIG_LOCKDEP 2005 unsigned char nested_level; 2006 #endif 2007 2008 2009 /* Protocol-specific pointers */ 2010 2011 #if IS_ENABLED(CONFIG_VLAN_8021Q) 2012 struct vlan_info __rcu *vlan_info; 2013 #endif 2014 #if IS_ENABLED(CONFIG_NET_DSA) 2015 struct dsa_port *dsa_ptr; 2016 #endif 2017 #if IS_ENABLED(CONFIG_TIPC) 2018 struct tipc_bearer __rcu *tipc_ptr; 2019 #endif 2020 #if IS_ENABLED(CONFIG_IRDA) || IS_ENABLED(CONFIG_ATALK) 2021 void *atalk_ptr; 2022 #endif 2023 struct in_device __rcu *ip_ptr; 2024 #if IS_ENABLED(CONFIG_DECNET) 2025 struct dn_dev __rcu *dn_ptr; 2026 #endif 2027 struct inet6_dev __rcu *ip6_ptr; 2028 #if IS_ENABLED(CONFIG_AX25) 2029 void *ax25_ptr; 2030 #endif 2031 struct wireless_dev *ieee80211_ptr; 2032 struct wpan_dev *ieee802154_ptr; 2033 #if IS_ENABLED(CONFIG_MPLS_ROUTING) 2034 struct mpls_dev __rcu *mpls_ptr; 2035 #endif 2036 2037 /* 2038 * Cache lines mostly used on receive path (including eth_type_trans()) 2039 */ 2040 /* Interface address info used in eth_type_trans() */ 2041 unsigned char *dev_addr; 2042 2043 struct netdev_rx_queue *_rx; 2044 unsigned int num_rx_queues; 2045 unsigned int real_num_rx_queues; 2046 2047 struct bpf_prog __rcu *xdp_prog; 2048 unsigned long gro_flush_timeout; 2049 int napi_defer_hard_irqs; 2050 rx_handler_func_t __rcu *rx_handler; 2051 void __rcu *rx_handler_data; 2052 2053 #ifdef CONFIG_NET_CLS_ACT 2054 struct mini_Qdisc __rcu *miniq_ingress; 2055 #endif 2056 struct netdev_queue __rcu *ingress_queue; 2057 #ifdef CONFIG_NETFILTER_INGRESS 2058 struct nf_hook_entries __rcu *nf_hooks_ingress; 2059 #endif 2060 2061 unsigned char broadcast[MAX_ADDR_LEN]; 2062 #ifdef CONFIG_RFS_ACCEL 2063 struct cpu_rmap *rx_cpu_rmap; 2064 #endif 2065 struct hlist_node index_hlist; 2066 2067 /* 2068 * Cache lines mostly used on transmit path 2069 */ 2070 struct netdev_queue *_tx ____cacheline_aligned_in_smp; 2071 unsigned int num_tx_queues; 2072 unsigned int real_num_tx_queues; 2073 struct Qdisc *qdisc; 2074 unsigned int tx_queue_len; 2075 spinlock_t tx_global_lock; 2076 2077 struct xdp_dev_bulk_queue __percpu *xdp_bulkq; 2078 2079 #ifdef CONFIG_XPS 2080 struct xps_dev_maps __rcu *xps_maps[XPS_MAPS_MAX]; 2081 #endif 2082 #ifdef CONFIG_NET_CLS_ACT 2083 struct mini_Qdisc __rcu *miniq_egress; 2084 #endif 2085 2086 #ifdef CONFIG_NET_SCHED 2087 DECLARE_HASHTABLE (qdisc_hash, 4); 2088 #endif 2089 /* These may be needed for future network-power-down code. */ 2090 struct timer_list watchdog_timer; 2091 int watchdog_timeo; 2092 2093 u32 proto_down_reason; 2094 2095 struct list_head todo_list; 2096 2097 #ifdef CONFIG_PCPU_DEV_REFCNT 2098 int __percpu *pcpu_refcnt; 2099 #else 2100 refcount_t dev_refcnt; 2101 #endif 2102 2103 struct list_head link_watch_list; 2104 2105 enum { NETREG_UNINITIALIZED=0, 2106 NETREG_REGISTERED, /* completed register_netdevice */ 2107 NETREG_UNREGISTERING, /* called unregister_netdevice */ 2108 NETREG_UNREGISTERED, /* completed unregister todo */ 2109 NETREG_RELEASED, /* called free_netdev */ 2110 NETREG_DUMMY, /* dummy device for NAPI poll */ 2111 } reg_state:8; 2112 2113 bool dismantle; 2114 2115 enum { 2116 RTNL_LINK_INITIALIZED, 2117 RTNL_LINK_INITIALIZING, 2118 } rtnl_link_state:16; 2119 2120 bool needs_free_netdev; 2121 void (*priv_destructor)(struct net_device *dev); 2122 2123 #ifdef CONFIG_NETPOLL 2124 struct netpoll_info __rcu *npinfo; 2125 #endif 2126 2127 possible_net_t nd_net; 2128 2129 /* mid-layer private */ 2130 void *ml_priv; 2131 enum netdev_ml_priv_type ml_priv_type; 2132 2133 union { 2134 struct pcpu_lstats __percpu *lstats; 2135 struct pcpu_sw_netstats __percpu *tstats; 2136 struct pcpu_dstats __percpu *dstats; 2137 }; 2138 2139 #if IS_ENABLED(CONFIG_GARP) 2140 struct garp_port __rcu *garp_port; 2141 #endif 2142 #if IS_ENABLED(CONFIG_MRP) 2143 struct mrp_port __rcu *mrp_port; 2144 #endif 2145 2146 struct device dev; 2147 const struct attribute_group *sysfs_groups[4]; 2148 const struct attribute_group *sysfs_rx_queue_group; 2149 2150 const struct rtnl_link_ops *rtnl_link_ops; 2151 2152 /* for setting kernel sock attribute on TCP connection setup */ 2153 #define GSO_MAX_SIZE 65536 2154 unsigned int gso_max_size; 2155 #define GSO_MAX_SEGS 65535 2156 u16 gso_max_segs; 2157 2158 #ifdef CONFIG_DCB 2159 const struct dcbnl_rtnl_ops *dcbnl_ops; 2160 #endif 2161 s16 num_tc; 2162 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE]; 2163 u8 prio_tc_map[TC_BITMASK + 1]; 2164 2165 #if IS_ENABLED(CONFIG_FCOE) 2166 unsigned int fcoe_ddp_xid; 2167 #endif 2168 #if IS_ENABLED(CONFIG_CGROUP_NET_PRIO) 2169 struct netprio_map __rcu *priomap; 2170 #endif 2171 struct phy_device *phydev; 2172 struct sfp_bus *sfp_bus; 2173 struct lock_class_key *qdisc_tx_busylock; 2174 struct lock_class_key *qdisc_running_key; 2175 bool proto_down; 2176 unsigned wol_enabled:1; 2177 unsigned threaded:1; 2178 2179 struct list_head net_notifier_list; 2180 2181 #if IS_ENABLED(CONFIG_MACSEC) 2182 /* MACsec management functions */ 2183 const struct macsec_ops *macsec_ops; 2184 #endif 2185 const struct udp_tunnel_nic_info *udp_tunnel_nic_info; 2186 struct udp_tunnel_nic *udp_tunnel_nic; 2187 2188 /* protected by rtnl_lock */ 2189 struct bpf_xdp_entity xdp_state[__MAX_XDP_MODE]; 2190 }; 2191 #define to_net_dev(d) container_of(d, struct net_device, dev) 2192 2193 static inline bool netif_elide_gro(const struct net_device *dev) 2194 { 2195 if (!(dev->features & NETIF_F_GRO) || dev->xdp_prog) 2196 return true; 2197 return false; 2198 } 2199 2200 #define NETDEV_ALIGN 32 2201 2202 static inline 2203 int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio) 2204 { 2205 return dev->prio_tc_map[prio & TC_BITMASK]; 2206 } 2207 2208 static inline 2209 int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc) 2210 { 2211 if (tc >= dev->num_tc) 2212 return -EINVAL; 2213 2214 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK; 2215 return 0; 2216 } 2217 2218 int netdev_txq_to_tc(struct net_device *dev, unsigned int txq); 2219 void netdev_reset_tc(struct net_device *dev); 2220 int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset); 2221 int netdev_set_num_tc(struct net_device *dev, u8 num_tc); 2222 2223 static inline 2224 int netdev_get_num_tc(struct net_device *dev) 2225 { 2226 return dev->num_tc; 2227 } 2228 2229 static inline void net_prefetch(void *p) 2230 { 2231 prefetch(p); 2232 #if L1_CACHE_BYTES < 128 2233 prefetch((u8 *)p + L1_CACHE_BYTES); 2234 #endif 2235 } 2236 2237 static inline void net_prefetchw(void *p) 2238 { 2239 prefetchw(p); 2240 #if L1_CACHE_BYTES < 128 2241 prefetchw((u8 *)p + L1_CACHE_BYTES); 2242 #endif 2243 } 2244 2245 void netdev_unbind_sb_channel(struct net_device *dev, 2246 struct net_device *sb_dev); 2247 int netdev_bind_sb_channel_queue(struct net_device *dev, 2248 struct net_device *sb_dev, 2249 u8 tc, u16 count, u16 offset); 2250 int netdev_set_sb_channel(struct net_device *dev, u16 channel); 2251 static inline int netdev_get_sb_channel(struct net_device *dev) 2252 { 2253 return max_t(int, -dev->num_tc, 0); 2254 } 2255 2256 static inline 2257 struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev, 2258 unsigned int index) 2259 { 2260 return &dev->_tx[index]; 2261 } 2262 2263 static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev, 2264 const struct sk_buff *skb) 2265 { 2266 return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb)); 2267 } 2268 2269 static inline void netdev_for_each_tx_queue(struct net_device *dev, 2270 void (*f)(struct net_device *, 2271 struct netdev_queue *, 2272 void *), 2273 void *arg) 2274 { 2275 unsigned int i; 2276 2277 for (i = 0; i < dev->num_tx_queues; i++) 2278 f(dev, &dev->_tx[i], arg); 2279 } 2280 2281 #define netdev_lockdep_set_classes(dev) \ 2282 { \ 2283 static struct lock_class_key qdisc_tx_busylock_key; \ 2284 static struct lock_class_key qdisc_running_key; \ 2285 static struct lock_class_key qdisc_xmit_lock_key; \ 2286 static struct lock_class_key dev_addr_list_lock_key; \ 2287 unsigned int i; \ 2288 \ 2289 (dev)->qdisc_tx_busylock = &qdisc_tx_busylock_key; \ 2290 (dev)->qdisc_running_key = &qdisc_running_key; \ 2291 lockdep_set_class(&(dev)->addr_list_lock, \ 2292 &dev_addr_list_lock_key); \ 2293 for (i = 0; i < (dev)->num_tx_queues; i++) \ 2294 lockdep_set_class(&(dev)->_tx[i]._xmit_lock, \ 2295 &qdisc_xmit_lock_key); \ 2296 } 2297 2298 u16 netdev_pick_tx(struct net_device *dev, struct sk_buff *skb, 2299 struct net_device *sb_dev); 2300 struct netdev_queue *netdev_core_pick_tx(struct net_device *dev, 2301 struct sk_buff *skb, 2302 struct net_device *sb_dev); 2303 2304 /* returns the headroom that the master device needs to take in account 2305 * when forwarding to this dev 2306 */ 2307 static inline unsigned netdev_get_fwd_headroom(struct net_device *dev) 2308 { 2309 return dev->priv_flags & IFF_PHONY_HEADROOM ? 0 : dev->needed_headroom; 2310 } 2311 2312 static inline void netdev_set_rx_headroom(struct net_device *dev, int new_hr) 2313 { 2314 if (dev->netdev_ops->ndo_set_rx_headroom) 2315 dev->netdev_ops->ndo_set_rx_headroom(dev, new_hr); 2316 } 2317 2318 /* set the device rx headroom to the dev's default */ 2319 static inline void netdev_reset_rx_headroom(struct net_device *dev) 2320 { 2321 netdev_set_rx_headroom(dev, -1); 2322 } 2323 2324 static inline void *netdev_get_ml_priv(struct net_device *dev, 2325 enum netdev_ml_priv_type type) 2326 { 2327 if (dev->ml_priv_type != type) 2328 return NULL; 2329 2330 return dev->ml_priv; 2331 } 2332 2333 static inline void netdev_set_ml_priv(struct net_device *dev, 2334 void *ml_priv, 2335 enum netdev_ml_priv_type type) 2336 { 2337 WARN(dev->ml_priv_type && dev->ml_priv_type != type, 2338 "Overwriting already set ml_priv_type (%u) with different ml_priv_type (%u)!\n", 2339 dev->ml_priv_type, type); 2340 WARN(!dev->ml_priv_type && dev->ml_priv, 2341 "Overwriting already set ml_priv and ml_priv_type is ML_PRIV_NONE!\n"); 2342 2343 dev->ml_priv = ml_priv; 2344 dev->ml_priv_type = type; 2345 } 2346 2347 /* 2348 * Net namespace inlines 2349 */ 2350 static inline 2351 struct net *dev_net(const struct net_device *dev) 2352 { 2353 return read_pnet(&dev->nd_net); 2354 } 2355 2356 static inline 2357 void dev_net_set(struct net_device *dev, struct net *net) 2358 { 2359 write_pnet(&dev->nd_net, net); 2360 } 2361 2362 /** 2363 * netdev_priv - access network device private data 2364 * @dev: network device 2365 * 2366 * Get network device private data 2367 */ 2368 static inline void *netdev_priv(const struct net_device *dev) 2369 { 2370 return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN); 2371 } 2372 2373 /* Set the sysfs physical device reference for the network logical device 2374 * if set prior to registration will cause a symlink during initialization. 2375 */ 2376 #define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev)) 2377 2378 /* Set the sysfs device type for the network logical device to allow 2379 * fine-grained identification of different network device types. For 2380 * example Ethernet, Wireless LAN, Bluetooth, WiMAX etc. 2381 */ 2382 #define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype)) 2383 2384 /* Default NAPI poll() weight 2385 * Device drivers are strongly advised to not use bigger value 2386 */ 2387 #define NAPI_POLL_WEIGHT 64 2388 2389 /** 2390 * netif_napi_add - initialize a NAPI context 2391 * @dev: network device 2392 * @napi: NAPI context 2393 * @poll: polling function 2394 * @weight: default weight 2395 * 2396 * netif_napi_add() must be used to initialize a NAPI context prior to calling 2397 * *any* of the other NAPI-related functions. 2398 */ 2399 void netif_napi_add(struct net_device *dev, struct napi_struct *napi, 2400 int (*poll)(struct napi_struct *, int), int weight); 2401 2402 /** 2403 * netif_tx_napi_add - initialize a NAPI context 2404 * @dev: network device 2405 * @napi: NAPI context 2406 * @poll: polling function 2407 * @weight: default weight 2408 * 2409 * This variant of netif_napi_add() should be used from drivers using NAPI 2410 * to exclusively poll a TX queue. 2411 * This will avoid we add it into napi_hash[], thus polluting this hash table. 2412 */ 2413 static inline void netif_tx_napi_add(struct net_device *dev, 2414 struct napi_struct *napi, 2415 int (*poll)(struct napi_struct *, int), 2416 int weight) 2417 { 2418 set_bit(NAPI_STATE_NO_BUSY_POLL, &napi->state); 2419 netif_napi_add(dev, napi, poll, weight); 2420 } 2421 2422 /** 2423 * __netif_napi_del - remove a NAPI context 2424 * @napi: NAPI context 2425 * 2426 * Warning: caller must observe RCU grace period before freeing memory 2427 * containing @napi. Drivers might want to call this helper to combine 2428 * all the needed RCU grace periods into a single one. 2429 */ 2430 void __netif_napi_del(struct napi_struct *napi); 2431 2432 /** 2433 * netif_napi_del - remove a NAPI context 2434 * @napi: NAPI context 2435 * 2436 * netif_napi_del() removes a NAPI context from the network device NAPI list 2437 */ 2438 static inline void netif_napi_del(struct napi_struct *napi) 2439 { 2440 __netif_napi_del(napi); 2441 synchronize_net(); 2442 } 2443 2444 struct napi_gro_cb { 2445 /* Virtual address of skb_shinfo(skb)->frags[0].page + offset. */ 2446 void *frag0; 2447 2448 /* Length of frag0. */ 2449 unsigned int frag0_len; 2450 2451 /* This indicates where we are processing relative to skb->data. */ 2452 int data_offset; 2453 2454 /* This is non-zero if the packet cannot be merged with the new skb. */ 2455 u16 flush; 2456 2457 /* Save the IP ID here and check when we get to the transport layer */ 2458 u16 flush_id; 2459 2460 /* Number of segments aggregated. */ 2461 u16 count; 2462 2463 /* Start offset for remote checksum offload */ 2464 u16 gro_remcsum_start; 2465 2466 /* jiffies when first packet was created/queued */ 2467 unsigned long age; 2468 2469 /* Used in ipv6_gro_receive() and foo-over-udp */ 2470 u16 proto; 2471 2472 /* This is non-zero if the packet may be of the same flow. */ 2473 u8 same_flow:1; 2474 2475 /* Used in tunnel GRO receive */ 2476 u8 encap_mark:1; 2477 2478 /* GRO checksum is valid */ 2479 u8 csum_valid:1; 2480 2481 /* Number of checksums via CHECKSUM_UNNECESSARY */ 2482 u8 csum_cnt:3; 2483 2484 /* Free the skb? */ 2485 u8 free:2; 2486 #define NAPI_GRO_FREE 1 2487 #define NAPI_GRO_FREE_STOLEN_HEAD 2 2488 2489 /* Used in foo-over-udp, set in udp[46]_gro_receive */ 2490 u8 is_ipv6:1; 2491 2492 /* Used in GRE, set in fou/gue_gro_receive */ 2493 u8 is_fou:1; 2494 2495 /* Used to determine if flush_id can be ignored */ 2496 u8 is_atomic:1; 2497 2498 /* Number of gro_receive callbacks this packet already went through */ 2499 u8 recursion_counter:4; 2500 2501 /* GRO is done by frag_list pointer chaining. */ 2502 u8 is_flist:1; 2503 2504 /* used to support CHECKSUM_COMPLETE for tunneling protocols */ 2505 __wsum csum; 2506 2507 /* used in skb_gro_receive() slow path */ 2508 struct sk_buff *last; 2509 }; 2510 2511 #define NAPI_GRO_CB(skb) ((struct napi_gro_cb *)(skb)->cb) 2512 2513 #define GRO_RECURSION_LIMIT 15 2514 static inline int gro_recursion_inc_test(struct sk_buff *skb) 2515 { 2516 return ++NAPI_GRO_CB(skb)->recursion_counter == GRO_RECURSION_LIMIT; 2517 } 2518 2519 typedef struct sk_buff *(*gro_receive_t)(struct list_head *, struct sk_buff *); 2520 static inline struct sk_buff *call_gro_receive(gro_receive_t cb, 2521 struct list_head *head, 2522 struct sk_buff *skb) 2523 { 2524 if (unlikely(gro_recursion_inc_test(skb))) { 2525 NAPI_GRO_CB(skb)->flush |= 1; 2526 return NULL; 2527 } 2528 2529 return cb(head, skb); 2530 } 2531 2532 typedef struct sk_buff *(*gro_receive_sk_t)(struct sock *, struct list_head *, 2533 struct sk_buff *); 2534 static inline struct sk_buff *call_gro_receive_sk(gro_receive_sk_t cb, 2535 struct sock *sk, 2536 struct list_head *head, 2537 struct sk_buff *skb) 2538 { 2539 if (unlikely(gro_recursion_inc_test(skb))) { 2540 NAPI_GRO_CB(skb)->flush |= 1; 2541 return NULL; 2542 } 2543 2544 return cb(sk, head, skb); 2545 } 2546 2547 struct packet_type { 2548 __be16 type; /* This is really htons(ether_type). */ 2549 bool ignore_outgoing; 2550 struct net_device *dev; /* NULL is wildcarded here */ 2551 int (*func) (struct sk_buff *, 2552 struct net_device *, 2553 struct packet_type *, 2554 struct net_device *); 2555 void (*list_func) (struct list_head *, 2556 struct packet_type *, 2557 struct net_device *); 2558 bool (*id_match)(struct packet_type *ptype, 2559 struct sock *sk); 2560 void *af_packet_priv; 2561 struct list_head list; 2562 }; 2563 2564 struct offload_callbacks { 2565 struct sk_buff *(*gso_segment)(struct sk_buff *skb, 2566 netdev_features_t features); 2567 struct sk_buff *(*gro_receive)(struct list_head *head, 2568 struct sk_buff *skb); 2569 int (*gro_complete)(struct sk_buff *skb, int nhoff); 2570 }; 2571 2572 struct packet_offload { 2573 __be16 type; /* This is really htons(ether_type). */ 2574 u16 priority; 2575 struct offload_callbacks callbacks; 2576 struct list_head list; 2577 }; 2578 2579 /* often modified stats are per-CPU, other are shared (netdev->stats) */ 2580 struct pcpu_sw_netstats { 2581 u64 rx_packets; 2582 u64 rx_bytes; 2583 u64 tx_packets; 2584 u64 tx_bytes; 2585 struct u64_stats_sync syncp; 2586 } __aligned(4 * sizeof(u64)); 2587 2588 struct pcpu_lstats { 2589 u64_stats_t packets; 2590 u64_stats_t bytes; 2591 struct u64_stats_sync syncp; 2592 } __aligned(2 * sizeof(u64)); 2593 2594 void dev_lstats_read(struct net_device *dev, u64 *packets, u64 *bytes); 2595 2596 static inline void dev_sw_netstats_rx_add(struct net_device *dev, unsigned int len) 2597 { 2598 struct pcpu_sw_netstats *tstats = this_cpu_ptr(dev->tstats); 2599 2600 u64_stats_update_begin(&tstats->syncp); 2601 tstats->rx_bytes += len; 2602 tstats->rx_packets++; 2603 u64_stats_update_end(&tstats->syncp); 2604 } 2605 2606 static inline void dev_sw_netstats_tx_add(struct net_device *dev, 2607 unsigned int packets, 2608 unsigned int len) 2609 { 2610 struct pcpu_sw_netstats *tstats = this_cpu_ptr(dev->tstats); 2611 2612 u64_stats_update_begin(&tstats->syncp); 2613 tstats->tx_bytes += len; 2614 tstats->tx_packets += packets; 2615 u64_stats_update_end(&tstats->syncp); 2616 } 2617 2618 static inline void dev_lstats_add(struct net_device *dev, unsigned int len) 2619 { 2620 struct pcpu_lstats *lstats = this_cpu_ptr(dev->lstats); 2621 2622 u64_stats_update_begin(&lstats->syncp); 2623 u64_stats_add(&lstats->bytes, len); 2624 u64_stats_inc(&lstats->packets); 2625 u64_stats_update_end(&lstats->syncp); 2626 } 2627 2628 #define __netdev_alloc_pcpu_stats(type, gfp) \ 2629 ({ \ 2630 typeof(type) __percpu *pcpu_stats = alloc_percpu_gfp(type, gfp);\ 2631 if (pcpu_stats) { \ 2632 int __cpu; \ 2633 for_each_possible_cpu(__cpu) { \ 2634 typeof(type) *stat; \ 2635 stat = per_cpu_ptr(pcpu_stats, __cpu); \ 2636 u64_stats_init(&stat->syncp); \ 2637 } \ 2638 } \ 2639 pcpu_stats; \ 2640 }) 2641 2642 #define netdev_alloc_pcpu_stats(type) \ 2643 __netdev_alloc_pcpu_stats(type, GFP_KERNEL) 2644 2645 #define devm_netdev_alloc_pcpu_stats(dev, type) \ 2646 ({ \ 2647 typeof(type) __percpu *pcpu_stats = devm_alloc_percpu(dev, type);\ 2648 if (pcpu_stats) { \ 2649 int __cpu; \ 2650 for_each_possible_cpu(__cpu) { \ 2651 typeof(type) *stat; \ 2652 stat = per_cpu_ptr(pcpu_stats, __cpu); \ 2653 u64_stats_init(&stat->syncp); \ 2654 } \ 2655 } \ 2656 pcpu_stats; \ 2657 }) 2658 2659 enum netdev_lag_tx_type { 2660 NETDEV_LAG_TX_TYPE_UNKNOWN, 2661 NETDEV_LAG_TX_TYPE_RANDOM, 2662 NETDEV_LAG_TX_TYPE_BROADCAST, 2663 NETDEV_LAG_TX_TYPE_ROUNDROBIN, 2664 NETDEV_LAG_TX_TYPE_ACTIVEBACKUP, 2665 NETDEV_LAG_TX_TYPE_HASH, 2666 }; 2667 2668 enum netdev_lag_hash { 2669 NETDEV_LAG_HASH_NONE, 2670 NETDEV_LAG_HASH_L2, 2671 NETDEV_LAG_HASH_L34, 2672 NETDEV_LAG_HASH_L23, 2673 NETDEV_LAG_HASH_E23, 2674 NETDEV_LAG_HASH_E34, 2675 NETDEV_LAG_HASH_VLAN_SRCMAC, 2676 NETDEV_LAG_HASH_UNKNOWN, 2677 }; 2678 2679 struct netdev_lag_upper_info { 2680 enum netdev_lag_tx_type tx_type; 2681 enum netdev_lag_hash hash_type; 2682 }; 2683 2684 struct netdev_lag_lower_state_info { 2685 u8 link_up : 1, 2686 tx_enabled : 1; 2687 }; 2688 2689 #include <linux/notifier.h> 2690 2691 /* netdevice notifier chain. Please remember to update netdev_cmd_to_name() 2692 * and the rtnetlink notification exclusion list in rtnetlink_event() when 2693 * adding new types. 2694 */ 2695 enum netdev_cmd { 2696 NETDEV_UP = 1, /* For now you can't veto a device up/down */ 2697 NETDEV_DOWN, 2698 NETDEV_REBOOT, /* Tell a protocol stack a network interface 2699 detected a hardware crash and restarted 2700 - we can use this eg to kick tcp sessions 2701 once done */ 2702 NETDEV_CHANGE, /* Notify device state change */ 2703 NETDEV_REGISTER, 2704 NETDEV_UNREGISTER, 2705 NETDEV_CHANGEMTU, /* notify after mtu change happened */ 2706 NETDEV_CHANGEADDR, /* notify after the address change */ 2707 NETDEV_PRE_CHANGEADDR, /* notify before the address change */ 2708 NETDEV_GOING_DOWN, 2709 NETDEV_CHANGENAME, 2710 NETDEV_FEAT_CHANGE, 2711 NETDEV_BONDING_FAILOVER, 2712 NETDEV_PRE_UP, 2713 NETDEV_PRE_TYPE_CHANGE, 2714 NETDEV_POST_TYPE_CHANGE, 2715 NETDEV_POST_INIT, 2716 NETDEV_RELEASE, 2717 NETDEV_NOTIFY_PEERS, 2718 NETDEV_JOIN, 2719 NETDEV_CHANGEUPPER, 2720 NETDEV_RESEND_IGMP, 2721 NETDEV_PRECHANGEMTU, /* notify before mtu change happened */ 2722 NETDEV_CHANGEINFODATA, 2723 NETDEV_BONDING_INFO, 2724 NETDEV_PRECHANGEUPPER, 2725 NETDEV_CHANGELOWERSTATE, 2726 NETDEV_UDP_TUNNEL_PUSH_INFO, 2727 NETDEV_UDP_TUNNEL_DROP_INFO, 2728 NETDEV_CHANGE_TX_QUEUE_LEN, 2729 NETDEV_CVLAN_FILTER_PUSH_INFO, 2730 NETDEV_CVLAN_FILTER_DROP_INFO, 2731 NETDEV_SVLAN_FILTER_PUSH_INFO, 2732 NETDEV_SVLAN_FILTER_DROP_INFO, 2733 }; 2734 const char *netdev_cmd_to_name(enum netdev_cmd cmd); 2735 2736 int register_netdevice_notifier(struct notifier_block *nb); 2737 int unregister_netdevice_notifier(struct notifier_block *nb); 2738 int register_netdevice_notifier_net(struct net *net, struct notifier_block *nb); 2739 int unregister_netdevice_notifier_net(struct net *net, 2740 struct notifier_block *nb); 2741 int register_netdevice_notifier_dev_net(struct net_device *dev, 2742 struct notifier_block *nb, 2743 struct netdev_net_notifier *nn); 2744 int unregister_netdevice_notifier_dev_net(struct net_device *dev, 2745 struct notifier_block *nb, 2746 struct netdev_net_notifier *nn); 2747 2748 struct netdev_notifier_info { 2749 struct net_device *dev; 2750 struct netlink_ext_ack *extack; 2751 }; 2752 2753 struct netdev_notifier_info_ext { 2754 struct netdev_notifier_info info; /* must be first */ 2755 union { 2756 u32 mtu; 2757 } ext; 2758 }; 2759 2760 struct netdev_notifier_change_info { 2761 struct netdev_notifier_info info; /* must be first */ 2762 unsigned int flags_changed; 2763 }; 2764 2765 struct netdev_notifier_changeupper_info { 2766 struct netdev_notifier_info info; /* must be first */ 2767 struct net_device *upper_dev; /* new upper dev */ 2768 bool master; /* is upper dev master */ 2769 bool linking; /* is the notification for link or unlink */ 2770 void *upper_info; /* upper dev info */ 2771 }; 2772 2773 struct netdev_notifier_changelowerstate_info { 2774 struct netdev_notifier_info info; /* must be first */ 2775 void *lower_state_info; /* is lower dev state */ 2776 }; 2777 2778 struct netdev_notifier_pre_changeaddr_info { 2779 struct netdev_notifier_info info; /* must be first */ 2780 const unsigned char *dev_addr; 2781 }; 2782 2783 static inline void netdev_notifier_info_init(struct netdev_notifier_info *info, 2784 struct net_device *dev) 2785 { 2786 info->dev = dev; 2787 info->extack = NULL; 2788 } 2789 2790 static inline struct net_device * 2791 netdev_notifier_info_to_dev(const struct netdev_notifier_info *info) 2792 { 2793 return info->dev; 2794 } 2795 2796 static inline struct netlink_ext_ack * 2797 netdev_notifier_info_to_extack(const struct netdev_notifier_info *info) 2798 { 2799 return info->extack; 2800 } 2801 2802 int call_netdevice_notifiers(unsigned long val, struct net_device *dev); 2803 2804 2805 extern rwlock_t dev_base_lock; /* Device list lock */ 2806 2807 #define for_each_netdev(net, d) \ 2808 list_for_each_entry(d, &(net)->dev_base_head, dev_list) 2809 #define for_each_netdev_reverse(net, d) \ 2810 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list) 2811 #define for_each_netdev_rcu(net, d) \ 2812 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list) 2813 #define for_each_netdev_safe(net, d, n) \ 2814 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list) 2815 #define for_each_netdev_continue(net, d) \ 2816 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list) 2817 #define for_each_netdev_continue_reverse(net, d) \ 2818 list_for_each_entry_continue_reverse(d, &(net)->dev_base_head, \ 2819 dev_list) 2820 #define for_each_netdev_continue_rcu(net, d) \ 2821 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list) 2822 #define for_each_netdev_in_bond_rcu(bond, slave) \ 2823 for_each_netdev_rcu(&init_net, slave) \ 2824 if (netdev_master_upper_dev_get_rcu(slave) == (bond)) 2825 #define net_device_entry(lh) list_entry(lh, struct net_device, dev_list) 2826 2827 static inline struct net_device *next_net_device(struct net_device *dev) 2828 { 2829 struct list_head *lh; 2830 struct net *net; 2831 2832 net = dev_net(dev); 2833 lh = dev->dev_list.next; 2834 return lh == &net->dev_base_head ? NULL : net_device_entry(lh); 2835 } 2836 2837 static inline struct net_device *next_net_device_rcu(struct net_device *dev) 2838 { 2839 struct list_head *lh; 2840 struct net *net; 2841 2842 net = dev_net(dev); 2843 lh = rcu_dereference(list_next_rcu(&dev->dev_list)); 2844 return lh == &net->dev_base_head ? NULL : net_device_entry(lh); 2845 } 2846 2847 static inline struct net_device *first_net_device(struct net *net) 2848 { 2849 return list_empty(&net->dev_base_head) ? NULL : 2850 net_device_entry(net->dev_base_head.next); 2851 } 2852 2853 static inline struct net_device *first_net_device_rcu(struct net *net) 2854 { 2855 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head)); 2856 2857 return lh == &net->dev_base_head ? NULL : net_device_entry(lh); 2858 } 2859 2860 int netdev_boot_setup_check(struct net_device *dev); 2861 unsigned long netdev_boot_base(const char *prefix, int unit); 2862 struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type, 2863 const char *hwaddr); 2864 struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type); 2865 void dev_add_pack(struct packet_type *pt); 2866 void dev_remove_pack(struct packet_type *pt); 2867 void __dev_remove_pack(struct packet_type *pt); 2868 void dev_add_offload(struct packet_offload *po); 2869 void dev_remove_offload(struct packet_offload *po); 2870 2871 int dev_get_iflink(const struct net_device *dev); 2872 int dev_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb); 2873 struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags, 2874 unsigned short mask); 2875 struct net_device *dev_get_by_name(struct net *net, const char *name); 2876 struct net_device *dev_get_by_name_rcu(struct net *net, const char *name); 2877 struct net_device *__dev_get_by_name(struct net *net, const char *name); 2878 int dev_alloc_name(struct net_device *dev, const char *name); 2879 int dev_open(struct net_device *dev, struct netlink_ext_ack *extack); 2880 void dev_close(struct net_device *dev); 2881 void dev_close_many(struct list_head *head, bool unlink); 2882 void dev_disable_lro(struct net_device *dev); 2883 int dev_loopback_xmit(struct net *net, struct sock *sk, struct sk_buff *newskb); 2884 u16 dev_pick_tx_zero(struct net_device *dev, struct sk_buff *skb, 2885 struct net_device *sb_dev); 2886 u16 dev_pick_tx_cpu_id(struct net_device *dev, struct sk_buff *skb, 2887 struct net_device *sb_dev); 2888 2889 int dev_queue_xmit(struct sk_buff *skb); 2890 int dev_queue_xmit_accel(struct sk_buff *skb, struct net_device *sb_dev); 2891 int __dev_direct_xmit(struct sk_buff *skb, u16 queue_id); 2892 2893 static inline int dev_direct_xmit(struct sk_buff *skb, u16 queue_id) 2894 { 2895 int ret; 2896 2897 ret = __dev_direct_xmit(skb, queue_id); 2898 if (!dev_xmit_complete(ret)) 2899 kfree_skb(skb); 2900 return ret; 2901 } 2902 2903 int register_netdevice(struct net_device *dev); 2904 void unregister_netdevice_queue(struct net_device *dev, struct list_head *head); 2905 void unregister_netdevice_many(struct list_head *head); 2906 static inline void unregister_netdevice(struct net_device *dev) 2907 { 2908 unregister_netdevice_queue(dev, NULL); 2909 } 2910 2911 int netdev_refcnt_read(const struct net_device *dev); 2912 void free_netdev(struct net_device *dev); 2913 void netdev_freemem(struct net_device *dev); 2914 int init_dummy_netdev(struct net_device *dev); 2915 2916 struct net_device *netdev_get_xmit_slave(struct net_device *dev, 2917 struct sk_buff *skb, 2918 bool all_slaves); 2919 struct net_device *netdev_sk_get_lowest_dev(struct net_device *dev, 2920 struct sock *sk); 2921 struct net_device *dev_get_by_index(struct net *net, int ifindex); 2922 struct net_device *__dev_get_by_index(struct net *net, int ifindex); 2923 struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex); 2924 struct net_device *dev_get_by_napi_id(unsigned int napi_id); 2925 int netdev_get_name(struct net *net, char *name, int ifindex); 2926 int dev_restart(struct net_device *dev); 2927 int skb_gro_receive(struct sk_buff *p, struct sk_buff *skb); 2928 int skb_gro_receive_list(struct sk_buff *p, struct sk_buff *skb); 2929 2930 static inline unsigned int skb_gro_offset(const struct sk_buff *skb) 2931 { 2932 return NAPI_GRO_CB(skb)->data_offset; 2933 } 2934 2935 static inline unsigned int skb_gro_len(const struct sk_buff *skb) 2936 { 2937 return skb->len - NAPI_GRO_CB(skb)->data_offset; 2938 } 2939 2940 static inline void skb_gro_pull(struct sk_buff *skb, unsigned int len) 2941 { 2942 NAPI_GRO_CB(skb)->data_offset += len; 2943 } 2944 2945 static inline void *skb_gro_header_fast(struct sk_buff *skb, 2946 unsigned int offset) 2947 { 2948 return NAPI_GRO_CB(skb)->frag0 + offset; 2949 } 2950 2951 static inline int skb_gro_header_hard(struct sk_buff *skb, unsigned int hlen) 2952 { 2953 return NAPI_GRO_CB(skb)->frag0_len < hlen; 2954 } 2955 2956 static inline void skb_gro_frag0_invalidate(struct sk_buff *skb) 2957 { 2958 NAPI_GRO_CB(skb)->frag0 = NULL; 2959 NAPI_GRO_CB(skb)->frag0_len = 0; 2960 } 2961 2962 static inline void *skb_gro_header_slow(struct sk_buff *skb, unsigned int hlen, 2963 unsigned int offset) 2964 { 2965 if (!pskb_may_pull(skb, hlen)) 2966 return NULL; 2967 2968 skb_gro_frag0_invalidate(skb); 2969 return skb->data + offset; 2970 } 2971 2972 static inline void *skb_gro_network_header(struct sk_buff *skb) 2973 { 2974 return (NAPI_GRO_CB(skb)->frag0 ?: skb->data) + 2975 skb_network_offset(skb); 2976 } 2977 2978 static inline void skb_gro_postpull_rcsum(struct sk_buff *skb, 2979 const void *start, unsigned int len) 2980 { 2981 if (NAPI_GRO_CB(skb)->csum_valid) 2982 NAPI_GRO_CB(skb)->csum = csum_sub(NAPI_GRO_CB(skb)->csum, 2983 csum_partial(start, len, 0)); 2984 } 2985 2986 /* GRO checksum functions. These are logical equivalents of the normal 2987 * checksum functions (in skbuff.h) except that they operate on the GRO 2988 * offsets and fields in sk_buff. 2989 */ 2990 2991 __sum16 __skb_gro_checksum_complete(struct sk_buff *skb); 2992 2993 static inline bool skb_at_gro_remcsum_start(struct sk_buff *skb) 2994 { 2995 return (NAPI_GRO_CB(skb)->gro_remcsum_start == skb_gro_offset(skb)); 2996 } 2997 2998 static inline bool __skb_gro_checksum_validate_needed(struct sk_buff *skb, 2999 bool zero_okay, 3000 __sum16 check) 3001 { 3002 return ((skb->ip_summed != CHECKSUM_PARTIAL || 3003 skb_checksum_start_offset(skb) < 3004 skb_gro_offset(skb)) && 3005 !skb_at_gro_remcsum_start(skb) && 3006 NAPI_GRO_CB(skb)->csum_cnt == 0 && 3007 (!zero_okay || check)); 3008 } 3009 3010 static inline __sum16 __skb_gro_checksum_validate_complete(struct sk_buff *skb, 3011 __wsum psum) 3012 { 3013 if (NAPI_GRO_CB(skb)->csum_valid && 3014 !csum_fold(csum_add(psum, NAPI_GRO_CB(skb)->csum))) 3015 return 0; 3016 3017 NAPI_GRO_CB(skb)->csum = psum; 3018 3019 return __skb_gro_checksum_complete(skb); 3020 } 3021 3022 static inline void skb_gro_incr_csum_unnecessary(struct sk_buff *skb) 3023 { 3024 if (NAPI_GRO_CB(skb)->csum_cnt > 0) { 3025 /* Consume a checksum from CHECKSUM_UNNECESSARY */ 3026 NAPI_GRO_CB(skb)->csum_cnt--; 3027 } else { 3028 /* Update skb for CHECKSUM_UNNECESSARY and csum_level when we 3029 * verified a new top level checksum or an encapsulated one 3030 * during GRO. This saves work if we fallback to normal path. 3031 */ 3032 __skb_incr_checksum_unnecessary(skb); 3033 } 3034 } 3035 3036 #define __skb_gro_checksum_validate(skb, proto, zero_okay, check, \ 3037 compute_pseudo) \ 3038 ({ \ 3039 __sum16 __ret = 0; \ 3040 if (__skb_gro_checksum_validate_needed(skb, zero_okay, check)) \ 3041 __ret = __skb_gro_checksum_validate_complete(skb, \ 3042 compute_pseudo(skb, proto)); \ 3043 if (!__ret) \ 3044 skb_gro_incr_csum_unnecessary(skb); \ 3045 __ret; \ 3046 }) 3047 3048 #define skb_gro_checksum_validate(skb, proto, compute_pseudo) \ 3049 __skb_gro_checksum_validate(skb, proto, false, 0, compute_pseudo) 3050 3051 #define skb_gro_checksum_validate_zero_check(skb, proto, check, \ 3052 compute_pseudo) \ 3053 __skb_gro_checksum_validate(skb, proto, true, check, compute_pseudo) 3054 3055 #define skb_gro_checksum_simple_validate(skb) \ 3056 __skb_gro_checksum_validate(skb, 0, false, 0, null_compute_pseudo) 3057 3058 static inline bool __skb_gro_checksum_convert_check(struct sk_buff *skb) 3059 { 3060 return (NAPI_GRO_CB(skb)->csum_cnt == 0 && 3061 !NAPI_GRO_CB(skb)->csum_valid); 3062 } 3063 3064 static inline void __skb_gro_checksum_convert(struct sk_buff *skb, 3065 __wsum pseudo) 3066 { 3067 NAPI_GRO_CB(skb)->csum = ~pseudo; 3068 NAPI_GRO_CB(skb)->csum_valid = 1; 3069 } 3070 3071 #define skb_gro_checksum_try_convert(skb, proto, compute_pseudo) \ 3072 do { \ 3073 if (__skb_gro_checksum_convert_check(skb)) \ 3074 __skb_gro_checksum_convert(skb, \ 3075 compute_pseudo(skb, proto)); \ 3076 } while (0) 3077 3078 struct gro_remcsum { 3079 int offset; 3080 __wsum delta; 3081 }; 3082 3083 static inline void skb_gro_remcsum_init(struct gro_remcsum *grc) 3084 { 3085 grc->offset = 0; 3086 grc->delta = 0; 3087 } 3088 3089 static inline void *skb_gro_remcsum_process(struct sk_buff *skb, void *ptr, 3090 unsigned int off, size_t hdrlen, 3091 int start, int offset, 3092 struct gro_remcsum *grc, 3093 bool nopartial) 3094 { 3095 __wsum delta; 3096 size_t plen = hdrlen + max_t(size_t, offset + sizeof(u16), start); 3097 3098 BUG_ON(!NAPI_GRO_CB(skb)->csum_valid); 3099 3100 if (!nopartial) { 3101 NAPI_GRO_CB(skb)->gro_remcsum_start = off + hdrlen + start; 3102 return ptr; 3103 } 3104 3105 ptr = skb_gro_header_fast(skb, off); 3106 if (skb_gro_header_hard(skb, off + plen)) { 3107 ptr = skb_gro_header_slow(skb, off + plen, off); 3108 if (!ptr) 3109 return NULL; 3110 } 3111 3112 delta = remcsum_adjust(ptr + hdrlen, NAPI_GRO_CB(skb)->csum, 3113 start, offset); 3114 3115 /* Adjust skb->csum since we changed the packet */ 3116 NAPI_GRO_CB(skb)->csum = csum_add(NAPI_GRO_CB(skb)->csum, delta); 3117 3118 grc->offset = off + hdrlen + offset; 3119 grc->delta = delta; 3120 3121 return ptr; 3122 } 3123 3124 static inline void skb_gro_remcsum_cleanup(struct sk_buff *skb, 3125 struct gro_remcsum *grc) 3126 { 3127 void *ptr; 3128 size_t plen = grc->offset + sizeof(u16); 3129 3130 if (!grc->delta) 3131 return; 3132 3133 ptr = skb_gro_header_fast(skb, grc->offset); 3134 if (skb_gro_header_hard(skb, grc->offset + sizeof(u16))) { 3135 ptr = skb_gro_header_slow(skb, plen, grc->offset); 3136 if (!ptr) 3137 return; 3138 } 3139 3140 remcsum_unadjust((__sum16 *)ptr, grc->delta); 3141 } 3142 3143 #ifdef CONFIG_XFRM_OFFLOAD 3144 static inline void skb_gro_flush_final(struct sk_buff *skb, struct sk_buff *pp, int flush) 3145 { 3146 if (PTR_ERR(pp) != -EINPROGRESS) 3147 NAPI_GRO_CB(skb)->flush |= flush; 3148 } 3149 static inline void skb_gro_flush_final_remcsum(struct sk_buff *skb, 3150 struct sk_buff *pp, 3151 int flush, 3152 struct gro_remcsum *grc) 3153 { 3154 if (PTR_ERR(pp) != -EINPROGRESS) { 3155 NAPI_GRO_CB(skb)->flush |= flush; 3156 skb_gro_remcsum_cleanup(skb, grc); 3157 skb->remcsum_offload = 0; 3158 } 3159 } 3160 #else 3161 static inline void skb_gro_flush_final(struct sk_buff *skb, struct sk_buff *pp, int flush) 3162 { 3163 NAPI_GRO_CB(skb)->flush |= flush; 3164 } 3165 static inline void skb_gro_flush_final_remcsum(struct sk_buff *skb, 3166 struct sk_buff *pp, 3167 int flush, 3168 struct gro_remcsum *grc) 3169 { 3170 NAPI_GRO_CB(skb)->flush |= flush; 3171 skb_gro_remcsum_cleanup(skb, grc); 3172 skb->remcsum_offload = 0; 3173 } 3174 #endif 3175 3176 static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev, 3177 unsigned short type, 3178 const void *daddr, const void *saddr, 3179 unsigned int len) 3180 { 3181 if (!dev->header_ops || !dev->header_ops->create) 3182 return 0; 3183 3184 return dev->header_ops->create(skb, dev, type, daddr, saddr, len); 3185 } 3186 3187 static inline int dev_parse_header(const struct sk_buff *skb, 3188 unsigned char *haddr) 3189 { 3190 const struct net_device *dev = skb->dev; 3191 3192 if (!dev->header_ops || !dev->header_ops->parse) 3193 return 0; 3194 return dev->header_ops->parse(skb, haddr); 3195 } 3196 3197 static inline __be16 dev_parse_header_protocol(const struct sk_buff *skb) 3198 { 3199 const struct net_device *dev = skb->dev; 3200 3201 if (!dev->header_ops || !dev->header_ops->parse_protocol) 3202 return 0; 3203 return dev->header_ops->parse_protocol(skb); 3204 } 3205 3206 /* ll_header must have at least hard_header_len allocated */ 3207 static inline bool dev_validate_header(const struct net_device *dev, 3208 char *ll_header, int len) 3209 { 3210 if (likely(len >= dev->hard_header_len)) 3211 return true; 3212 if (len < dev->min_header_len) 3213 return false; 3214 3215 if (capable(CAP_SYS_RAWIO)) { 3216 memset(ll_header + len, 0, dev->hard_header_len - len); 3217 return true; 3218 } 3219 3220 if (dev->header_ops && dev->header_ops->validate) 3221 return dev->header_ops->validate(ll_header, len); 3222 3223 return false; 3224 } 3225 3226 static inline bool dev_has_header(const struct net_device *dev) 3227 { 3228 return dev->header_ops && dev->header_ops->create; 3229 } 3230 3231 typedef int gifconf_func_t(struct net_device * dev, char __user * bufptr, 3232 int len, int size); 3233 int register_gifconf(unsigned int family, gifconf_func_t *gifconf); 3234 static inline int unregister_gifconf(unsigned int family) 3235 { 3236 return register_gifconf(family, NULL); 3237 } 3238 3239 #ifdef CONFIG_NET_FLOW_LIMIT 3240 #define FLOW_LIMIT_HISTORY (1 << 7) /* must be ^2 and !overflow buckets */ 3241 struct sd_flow_limit { 3242 u64 count; 3243 unsigned int num_buckets; 3244 unsigned int history_head; 3245 u16 history[FLOW_LIMIT_HISTORY]; 3246 u8 buckets[]; 3247 }; 3248 3249 extern int netdev_flow_limit_table_len; 3250 #endif /* CONFIG_NET_FLOW_LIMIT */ 3251 3252 /* 3253 * Incoming packets are placed on per-CPU queues 3254 */ 3255 struct softnet_data { 3256 struct list_head poll_list; 3257 struct sk_buff_head process_queue; 3258 3259 /* stats */ 3260 unsigned int processed; 3261 unsigned int time_squeeze; 3262 unsigned int received_rps; 3263 #ifdef CONFIG_RPS 3264 struct softnet_data *rps_ipi_list; 3265 #endif 3266 #ifdef CONFIG_NET_FLOW_LIMIT 3267 struct sd_flow_limit __rcu *flow_limit; 3268 #endif 3269 struct Qdisc *output_queue; 3270 struct Qdisc **output_queue_tailp; 3271 struct sk_buff *completion_queue; 3272 #ifdef CONFIG_XFRM_OFFLOAD 3273 struct sk_buff_head xfrm_backlog; 3274 #endif 3275 /* written and read only by owning cpu: */ 3276 struct { 3277 u16 recursion; 3278 u8 more; 3279 } xmit; 3280 #ifdef CONFIG_RPS 3281 /* input_queue_head should be written by cpu owning this struct, 3282 * and only read by other cpus. Worth using a cache line. 3283 */ 3284 unsigned int input_queue_head ____cacheline_aligned_in_smp; 3285 3286 /* Elements below can be accessed between CPUs for RPS/RFS */ 3287 call_single_data_t csd ____cacheline_aligned_in_smp; 3288 struct softnet_data *rps_ipi_next; 3289 unsigned int cpu; 3290 unsigned int input_queue_tail; 3291 #endif 3292 unsigned int dropped; 3293 struct sk_buff_head input_pkt_queue; 3294 struct napi_struct backlog; 3295 3296 }; 3297 3298 static inline void input_queue_head_incr(struct softnet_data *sd) 3299 { 3300 #ifdef CONFIG_RPS 3301 sd->input_queue_head++; 3302 #endif 3303 } 3304 3305 static inline void input_queue_tail_incr_save(struct softnet_data *sd, 3306 unsigned int *qtail) 3307 { 3308 #ifdef CONFIG_RPS 3309 *qtail = ++sd->input_queue_tail; 3310 #endif 3311 } 3312 3313 DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data); 3314 3315 static inline int dev_recursion_level(void) 3316 { 3317 return this_cpu_read(softnet_data.xmit.recursion); 3318 } 3319 3320 #define XMIT_RECURSION_LIMIT 8 3321 static inline bool dev_xmit_recursion(void) 3322 { 3323 return unlikely(__this_cpu_read(softnet_data.xmit.recursion) > 3324 XMIT_RECURSION_LIMIT); 3325 } 3326 3327 static inline void dev_xmit_recursion_inc(void) 3328 { 3329 __this_cpu_inc(softnet_data.xmit.recursion); 3330 } 3331 3332 static inline void dev_xmit_recursion_dec(void) 3333 { 3334 __this_cpu_dec(softnet_data.xmit.recursion); 3335 } 3336 3337 void __netif_schedule(struct Qdisc *q); 3338 void netif_schedule_queue(struct netdev_queue *txq); 3339 3340 static inline void netif_tx_schedule_all(struct net_device *dev) 3341 { 3342 unsigned int i; 3343 3344 for (i = 0; i < dev->num_tx_queues; i++) 3345 netif_schedule_queue(netdev_get_tx_queue(dev, i)); 3346 } 3347 3348 static __always_inline void netif_tx_start_queue(struct netdev_queue *dev_queue) 3349 { 3350 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state); 3351 } 3352 3353 /** 3354 * netif_start_queue - allow transmit 3355 * @dev: network device 3356 * 3357 * Allow upper layers to call the device hard_start_xmit routine. 3358 */ 3359 static inline void netif_start_queue(struct net_device *dev) 3360 { 3361 netif_tx_start_queue(netdev_get_tx_queue(dev, 0)); 3362 } 3363 3364 static inline void netif_tx_start_all_queues(struct net_device *dev) 3365 { 3366 unsigned int i; 3367 3368 for (i = 0; i < dev->num_tx_queues; i++) { 3369 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 3370 netif_tx_start_queue(txq); 3371 } 3372 } 3373 3374 void netif_tx_wake_queue(struct netdev_queue *dev_queue); 3375 3376 /** 3377 * netif_wake_queue - restart transmit 3378 * @dev: network device 3379 * 3380 * Allow upper layers to call the device hard_start_xmit routine. 3381 * Used for flow control when transmit resources are available. 3382 */ 3383 static inline void netif_wake_queue(struct net_device *dev) 3384 { 3385 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0)); 3386 } 3387 3388 static inline void netif_tx_wake_all_queues(struct net_device *dev) 3389 { 3390 unsigned int i; 3391 3392 for (i = 0; i < dev->num_tx_queues; i++) { 3393 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 3394 netif_tx_wake_queue(txq); 3395 } 3396 } 3397 3398 static __always_inline void netif_tx_stop_queue(struct netdev_queue *dev_queue) 3399 { 3400 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state); 3401 } 3402 3403 /** 3404 * netif_stop_queue - stop transmitted packets 3405 * @dev: network device 3406 * 3407 * Stop upper layers calling the device hard_start_xmit routine. 3408 * Used for flow control when transmit resources are unavailable. 3409 */ 3410 static inline void netif_stop_queue(struct net_device *dev) 3411 { 3412 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0)); 3413 } 3414 3415 void netif_tx_stop_all_queues(struct net_device *dev); 3416 3417 static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue) 3418 { 3419 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state); 3420 } 3421 3422 /** 3423 * netif_queue_stopped - test if transmit queue is flowblocked 3424 * @dev: network device 3425 * 3426 * Test if transmit queue on device is currently unable to send. 3427 */ 3428 static inline bool netif_queue_stopped(const struct net_device *dev) 3429 { 3430 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0)); 3431 } 3432 3433 static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue) 3434 { 3435 return dev_queue->state & QUEUE_STATE_ANY_XOFF; 3436 } 3437 3438 static inline bool 3439 netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue) 3440 { 3441 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN; 3442 } 3443 3444 static inline bool 3445 netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue) 3446 { 3447 return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN; 3448 } 3449 3450 /** 3451 * netdev_queue_set_dql_min_limit - set dql minimum limit 3452 * @dev_queue: pointer to transmit queue 3453 * @min_limit: dql minimum limit 3454 * 3455 * Forces xmit_more() to return true until the minimum threshold 3456 * defined by @min_limit is reached (or until the tx queue is 3457 * empty). Warning: to be use with care, misuse will impact the 3458 * latency. 3459 */ 3460 static inline void netdev_queue_set_dql_min_limit(struct netdev_queue *dev_queue, 3461 unsigned int min_limit) 3462 { 3463 #ifdef CONFIG_BQL 3464 dev_queue->dql.min_limit = min_limit; 3465 #endif 3466 } 3467 3468 /** 3469 * netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write 3470 * @dev_queue: pointer to transmit queue 3471 * 3472 * BQL enabled drivers might use this helper in their ndo_start_xmit(), 3473 * to give appropriate hint to the CPU. 3474 */ 3475 static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue) 3476 { 3477 #ifdef CONFIG_BQL 3478 prefetchw(&dev_queue->dql.num_queued); 3479 #endif 3480 } 3481 3482 /** 3483 * netdev_txq_bql_complete_prefetchw - prefetch bql data for write 3484 * @dev_queue: pointer to transmit queue 3485 * 3486 * BQL enabled drivers might use this helper in their TX completion path, 3487 * to give appropriate hint to the CPU. 3488 */ 3489 static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue) 3490 { 3491 #ifdef CONFIG_BQL 3492 prefetchw(&dev_queue->dql.limit); 3493 #endif 3494 } 3495 3496 static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue, 3497 unsigned int bytes) 3498 { 3499 #ifdef CONFIG_BQL 3500 dql_queued(&dev_queue->dql, bytes); 3501 3502 if (likely(dql_avail(&dev_queue->dql) >= 0)) 3503 return; 3504 3505 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state); 3506 3507 /* 3508 * The XOFF flag must be set before checking the dql_avail below, 3509 * because in netdev_tx_completed_queue we update the dql_completed 3510 * before checking the XOFF flag. 3511 */ 3512 smp_mb(); 3513 3514 /* check again in case another CPU has just made room avail */ 3515 if (unlikely(dql_avail(&dev_queue->dql) >= 0)) 3516 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state); 3517 #endif 3518 } 3519 3520 /* Variant of netdev_tx_sent_queue() for drivers that are aware 3521 * that they should not test BQL status themselves. 3522 * We do want to change __QUEUE_STATE_STACK_XOFF only for the last 3523 * skb of a batch. 3524 * Returns true if the doorbell must be used to kick the NIC. 3525 */ 3526 static inline bool __netdev_tx_sent_queue(struct netdev_queue *dev_queue, 3527 unsigned int bytes, 3528 bool xmit_more) 3529 { 3530 if (xmit_more) { 3531 #ifdef CONFIG_BQL 3532 dql_queued(&dev_queue->dql, bytes); 3533 #endif 3534 return netif_tx_queue_stopped(dev_queue); 3535 } 3536 netdev_tx_sent_queue(dev_queue, bytes); 3537 return true; 3538 } 3539 3540 /** 3541 * netdev_sent_queue - report the number of bytes queued to hardware 3542 * @dev: network device 3543 * @bytes: number of bytes queued to the hardware device queue 3544 * 3545 * Report the number of bytes queued for sending/completion to the network 3546 * device hardware queue. @bytes should be a good approximation and should 3547 * exactly match netdev_completed_queue() @bytes 3548 */ 3549 static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes) 3550 { 3551 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes); 3552 } 3553 3554 static inline bool __netdev_sent_queue(struct net_device *dev, 3555 unsigned int bytes, 3556 bool xmit_more) 3557 { 3558 return __netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes, 3559 xmit_more); 3560 } 3561 3562 static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue, 3563 unsigned int pkts, unsigned int bytes) 3564 { 3565 #ifdef CONFIG_BQL 3566 if (unlikely(!bytes)) 3567 return; 3568 3569 dql_completed(&dev_queue->dql, bytes); 3570 3571 /* 3572 * Without the memory barrier there is a small possiblity that 3573 * netdev_tx_sent_queue will miss the update and cause the queue to 3574 * be stopped forever 3575 */ 3576 smp_mb(); 3577 3578 if (unlikely(dql_avail(&dev_queue->dql) < 0)) 3579 return; 3580 3581 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state)) 3582 netif_schedule_queue(dev_queue); 3583 #endif 3584 } 3585 3586 /** 3587 * netdev_completed_queue - report bytes and packets completed by device 3588 * @dev: network device 3589 * @pkts: actual number of packets sent over the medium 3590 * @bytes: actual number of bytes sent over the medium 3591 * 3592 * Report the number of bytes and packets transmitted by the network device 3593 * hardware queue over the physical medium, @bytes must exactly match the 3594 * @bytes amount passed to netdev_sent_queue() 3595 */ 3596 static inline void netdev_completed_queue(struct net_device *dev, 3597 unsigned int pkts, unsigned int bytes) 3598 { 3599 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes); 3600 } 3601 3602 static inline void netdev_tx_reset_queue(struct netdev_queue *q) 3603 { 3604 #ifdef CONFIG_BQL 3605 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state); 3606 dql_reset(&q->dql); 3607 #endif 3608 } 3609 3610 /** 3611 * netdev_reset_queue - reset the packets and bytes count of a network device 3612 * @dev_queue: network device 3613 * 3614 * Reset the bytes and packet count of a network device and clear the 3615 * software flow control OFF bit for this network device 3616 */ 3617 static inline void netdev_reset_queue(struct net_device *dev_queue) 3618 { 3619 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0)); 3620 } 3621 3622 /** 3623 * netdev_cap_txqueue - check if selected tx queue exceeds device queues 3624 * @dev: network device 3625 * @queue_index: given tx queue index 3626 * 3627 * Returns 0 if given tx queue index >= number of device tx queues, 3628 * otherwise returns the originally passed tx queue index. 3629 */ 3630 static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index) 3631 { 3632 if (unlikely(queue_index >= dev->real_num_tx_queues)) { 3633 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n", 3634 dev->name, queue_index, 3635 dev->real_num_tx_queues); 3636 return 0; 3637 } 3638 3639 return queue_index; 3640 } 3641 3642 /** 3643 * netif_running - test if up 3644 * @dev: network device 3645 * 3646 * Test if the device has been brought up. 3647 */ 3648 static inline bool netif_running(const struct net_device *dev) 3649 { 3650 return test_bit(__LINK_STATE_START, &dev->state); 3651 } 3652 3653 /* 3654 * Routines to manage the subqueues on a device. We only need start, 3655 * stop, and a check if it's stopped. All other device management is 3656 * done at the overall netdevice level. 3657 * Also test the device if we're multiqueue. 3658 */ 3659 3660 /** 3661 * netif_start_subqueue - allow sending packets on subqueue 3662 * @dev: network device 3663 * @queue_index: sub queue index 3664 * 3665 * Start individual transmit queue of a device with multiple transmit queues. 3666 */ 3667 static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index) 3668 { 3669 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 3670 3671 netif_tx_start_queue(txq); 3672 } 3673 3674 /** 3675 * netif_stop_subqueue - stop sending packets on subqueue 3676 * @dev: network device 3677 * @queue_index: sub queue index 3678 * 3679 * Stop individual transmit queue of a device with multiple transmit queues. 3680 */ 3681 static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index) 3682 { 3683 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 3684 netif_tx_stop_queue(txq); 3685 } 3686 3687 /** 3688 * __netif_subqueue_stopped - test status of subqueue 3689 * @dev: network device 3690 * @queue_index: sub queue index 3691 * 3692 * Check individual transmit queue of a device with multiple transmit queues. 3693 */ 3694 static inline bool __netif_subqueue_stopped(const struct net_device *dev, 3695 u16 queue_index) 3696 { 3697 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 3698 3699 return netif_tx_queue_stopped(txq); 3700 } 3701 3702 /** 3703 * netif_subqueue_stopped - test status of subqueue 3704 * @dev: network device 3705 * @skb: sub queue buffer pointer 3706 * 3707 * Check individual transmit queue of a device with multiple transmit queues. 3708 */ 3709 static inline bool netif_subqueue_stopped(const struct net_device *dev, 3710 struct sk_buff *skb) 3711 { 3712 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb)); 3713 } 3714 3715 /** 3716 * netif_wake_subqueue - allow sending packets on subqueue 3717 * @dev: network device 3718 * @queue_index: sub queue index 3719 * 3720 * Resume individual transmit queue of a device with multiple transmit queues. 3721 */ 3722 static inline void netif_wake_subqueue(struct net_device *dev, u16 queue_index) 3723 { 3724 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 3725 3726 netif_tx_wake_queue(txq); 3727 } 3728 3729 #ifdef CONFIG_XPS 3730 int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask, 3731 u16 index); 3732 int __netif_set_xps_queue(struct net_device *dev, const unsigned long *mask, 3733 u16 index, enum xps_map_type type); 3734 3735 /** 3736 * netif_attr_test_mask - Test a CPU or Rx queue set in a mask 3737 * @j: CPU/Rx queue index 3738 * @mask: bitmask of all cpus/rx queues 3739 * @nr_bits: number of bits in the bitmask 3740 * 3741 * Test if a CPU or Rx queue index is set in a mask of all CPU/Rx queues. 3742 */ 3743 static inline bool netif_attr_test_mask(unsigned long j, 3744 const unsigned long *mask, 3745 unsigned int nr_bits) 3746 { 3747 cpu_max_bits_warn(j, nr_bits); 3748 return test_bit(j, mask); 3749 } 3750 3751 /** 3752 * netif_attr_test_online - Test for online CPU/Rx queue 3753 * @j: CPU/Rx queue index 3754 * @online_mask: bitmask for CPUs/Rx queues that are online 3755 * @nr_bits: number of bits in the bitmask 3756 * 3757 * Returns true if a CPU/Rx queue is online. 3758 */ 3759 static inline bool netif_attr_test_online(unsigned long j, 3760 const unsigned long *online_mask, 3761 unsigned int nr_bits) 3762 { 3763 cpu_max_bits_warn(j, nr_bits); 3764 3765 if (online_mask) 3766 return test_bit(j, online_mask); 3767 3768 return (j < nr_bits); 3769 } 3770 3771 /** 3772 * netif_attrmask_next - get the next CPU/Rx queue in a cpu/Rx queues mask 3773 * @n: CPU/Rx queue index 3774 * @srcp: the cpumask/Rx queue mask pointer 3775 * @nr_bits: number of bits in the bitmask 3776 * 3777 * Returns >= nr_bits if no further CPUs/Rx queues set. 3778 */ 3779 static inline unsigned int netif_attrmask_next(int n, const unsigned long *srcp, 3780 unsigned int nr_bits) 3781 { 3782 /* -1 is a legal arg here. */ 3783 if (n != -1) 3784 cpu_max_bits_warn(n, nr_bits); 3785 3786 if (srcp) 3787 return find_next_bit(srcp, nr_bits, n + 1); 3788 3789 return n + 1; 3790 } 3791 3792 /** 3793 * netif_attrmask_next_and - get the next CPU/Rx queue in \*src1p & \*src2p 3794 * @n: CPU/Rx queue index 3795 * @src1p: the first CPUs/Rx queues mask pointer 3796 * @src2p: the second CPUs/Rx queues mask pointer 3797 * @nr_bits: number of bits in the bitmask 3798 * 3799 * Returns >= nr_bits if no further CPUs/Rx queues set in both. 3800 */ 3801 static inline int netif_attrmask_next_and(int n, const unsigned long *src1p, 3802 const unsigned long *src2p, 3803 unsigned int nr_bits) 3804 { 3805 /* -1 is a legal arg here. */ 3806 if (n != -1) 3807 cpu_max_bits_warn(n, nr_bits); 3808 3809 if (src1p && src2p) 3810 return find_next_and_bit(src1p, src2p, nr_bits, n + 1); 3811 else if (src1p) 3812 return find_next_bit(src1p, nr_bits, n + 1); 3813 else if (src2p) 3814 return find_next_bit(src2p, nr_bits, n + 1); 3815 3816 return n + 1; 3817 } 3818 #else 3819 static inline int netif_set_xps_queue(struct net_device *dev, 3820 const struct cpumask *mask, 3821 u16 index) 3822 { 3823 return 0; 3824 } 3825 3826 static inline int __netif_set_xps_queue(struct net_device *dev, 3827 const unsigned long *mask, 3828 u16 index, enum xps_map_type type) 3829 { 3830 return 0; 3831 } 3832 #endif 3833 3834 /** 3835 * netif_is_multiqueue - test if device has multiple transmit queues 3836 * @dev: network device 3837 * 3838 * Check if device has multiple transmit queues 3839 */ 3840 static inline bool netif_is_multiqueue(const struct net_device *dev) 3841 { 3842 return dev->num_tx_queues > 1; 3843 } 3844 3845 int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq); 3846 3847 #ifdef CONFIG_SYSFS 3848 int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq); 3849 #else 3850 static inline int netif_set_real_num_rx_queues(struct net_device *dev, 3851 unsigned int rxqs) 3852 { 3853 dev->real_num_rx_queues = rxqs; 3854 return 0; 3855 } 3856 #endif 3857 3858 static inline struct netdev_rx_queue * 3859 __netif_get_rx_queue(struct net_device *dev, unsigned int rxq) 3860 { 3861 return dev->_rx + rxq; 3862 } 3863 3864 #ifdef CONFIG_SYSFS 3865 static inline unsigned int get_netdev_rx_queue_index( 3866 struct netdev_rx_queue *queue) 3867 { 3868 struct net_device *dev = queue->dev; 3869 int index = queue - dev->_rx; 3870 3871 BUG_ON(index >= dev->num_rx_queues); 3872 return index; 3873 } 3874 #endif 3875 3876 #define DEFAULT_MAX_NUM_RSS_QUEUES (8) 3877 int netif_get_num_default_rss_queues(void); 3878 3879 enum skb_free_reason { 3880 SKB_REASON_CONSUMED, 3881 SKB_REASON_DROPPED, 3882 }; 3883 3884 void __dev_kfree_skb_irq(struct sk_buff *skb, enum skb_free_reason reason); 3885 void __dev_kfree_skb_any(struct sk_buff *skb, enum skb_free_reason reason); 3886 3887 /* 3888 * It is not allowed to call kfree_skb() or consume_skb() from hardware 3889 * interrupt context or with hardware interrupts being disabled. 3890 * (in_irq() || irqs_disabled()) 3891 * 3892 * We provide four helpers that can be used in following contexts : 3893 * 3894 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context, 3895 * replacing kfree_skb(skb) 3896 * 3897 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context. 3898 * Typically used in place of consume_skb(skb) in TX completion path 3899 * 3900 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context, 3901 * replacing kfree_skb(skb) 3902 * 3903 * dev_consume_skb_any(skb) when caller doesn't know its current irq context, 3904 * and consumed a packet. Used in place of consume_skb(skb) 3905 */ 3906 static inline void dev_kfree_skb_irq(struct sk_buff *skb) 3907 { 3908 __dev_kfree_skb_irq(skb, SKB_REASON_DROPPED); 3909 } 3910 3911 static inline void dev_consume_skb_irq(struct sk_buff *skb) 3912 { 3913 __dev_kfree_skb_irq(skb, SKB_REASON_CONSUMED); 3914 } 3915 3916 static inline void dev_kfree_skb_any(struct sk_buff *skb) 3917 { 3918 __dev_kfree_skb_any(skb, SKB_REASON_DROPPED); 3919 } 3920 3921 static inline void dev_consume_skb_any(struct sk_buff *skb) 3922 { 3923 __dev_kfree_skb_any(skb, SKB_REASON_CONSUMED); 3924 } 3925 3926 void generic_xdp_tx(struct sk_buff *skb, struct bpf_prog *xdp_prog); 3927 int do_xdp_generic(struct bpf_prog *xdp_prog, struct sk_buff *skb); 3928 int netif_rx(struct sk_buff *skb); 3929 int netif_rx_ni(struct sk_buff *skb); 3930 int netif_rx_any_context(struct sk_buff *skb); 3931 int netif_receive_skb(struct sk_buff *skb); 3932 int netif_receive_skb_core(struct sk_buff *skb); 3933 void netif_receive_skb_list(struct list_head *head); 3934 gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb); 3935 void napi_gro_flush(struct napi_struct *napi, bool flush_old); 3936 struct sk_buff *napi_get_frags(struct napi_struct *napi); 3937 gro_result_t napi_gro_frags(struct napi_struct *napi); 3938 struct packet_offload *gro_find_receive_by_type(__be16 type); 3939 struct packet_offload *gro_find_complete_by_type(__be16 type); 3940 3941 static inline void napi_free_frags(struct napi_struct *napi) 3942 { 3943 kfree_skb(napi->skb); 3944 napi->skb = NULL; 3945 } 3946 3947 bool netdev_is_rx_handler_busy(struct net_device *dev); 3948 int netdev_rx_handler_register(struct net_device *dev, 3949 rx_handler_func_t *rx_handler, 3950 void *rx_handler_data); 3951 void netdev_rx_handler_unregister(struct net_device *dev); 3952 3953 bool dev_valid_name(const char *name); 3954 int dev_ioctl(struct net *net, unsigned int cmd, struct ifreq *ifr, 3955 bool *need_copyout); 3956 int dev_ifconf(struct net *net, struct ifconf *, int); 3957 int dev_ethtool(struct net *net, struct ifreq *); 3958 unsigned int dev_get_flags(const struct net_device *); 3959 int __dev_change_flags(struct net_device *dev, unsigned int flags, 3960 struct netlink_ext_ack *extack); 3961 int dev_change_flags(struct net_device *dev, unsigned int flags, 3962 struct netlink_ext_ack *extack); 3963 void __dev_notify_flags(struct net_device *, unsigned int old_flags, 3964 unsigned int gchanges); 3965 int dev_change_name(struct net_device *, const char *); 3966 int dev_set_alias(struct net_device *, const char *, size_t); 3967 int dev_get_alias(const struct net_device *, char *, size_t); 3968 int dev_change_net_namespace(struct net_device *, struct net *, const char *); 3969 int __dev_set_mtu(struct net_device *, int); 3970 int dev_validate_mtu(struct net_device *dev, int mtu, 3971 struct netlink_ext_ack *extack); 3972 int dev_set_mtu_ext(struct net_device *dev, int mtu, 3973 struct netlink_ext_ack *extack); 3974 int dev_set_mtu(struct net_device *, int); 3975 int dev_change_tx_queue_len(struct net_device *, unsigned long); 3976 void dev_set_group(struct net_device *, int); 3977 int dev_pre_changeaddr_notify(struct net_device *dev, const char *addr, 3978 struct netlink_ext_ack *extack); 3979 int dev_set_mac_address(struct net_device *dev, struct sockaddr *sa, 3980 struct netlink_ext_ack *extack); 3981 int dev_set_mac_address_user(struct net_device *dev, struct sockaddr *sa, 3982 struct netlink_ext_ack *extack); 3983 int dev_get_mac_address(struct sockaddr *sa, struct net *net, char *dev_name); 3984 int dev_change_carrier(struct net_device *, bool new_carrier); 3985 int dev_get_phys_port_id(struct net_device *dev, 3986 struct netdev_phys_item_id *ppid); 3987 int dev_get_phys_port_name(struct net_device *dev, 3988 char *name, size_t len); 3989 int dev_get_port_parent_id(struct net_device *dev, 3990 struct netdev_phys_item_id *ppid, bool recurse); 3991 bool netdev_port_same_parent_id(struct net_device *a, struct net_device *b); 3992 int dev_change_proto_down(struct net_device *dev, bool proto_down); 3993 int dev_change_proto_down_generic(struct net_device *dev, bool proto_down); 3994 void dev_change_proto_down_reason(struct net_device *dev, unsigned long mask, 3995 u32 value); 3996 struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev, bool *again); 3997 struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev, 3998 struct netdev_queue *txq, int *ret); 3999 4000 typedef int (*bpf_op_t)(struct net_device *dev, struct netdev_bpf *bpf); 4001 int dev_change_xdp_fd(struct net_device *dev, struct netlink_ext_ack *extack, 4002 int fd, int expected_fd, u32 flags); 4003 int bpf_xdp_link_attach(const union bpf_attr *attr, struct bpf_prog *prog); 4004 u32 dev_xdp_prog_id(struct net_device *dev, enum bpf_xdp_mode mode); 4005 4006 int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb); 4007 int dev_forward_skb(struct net_device *dev, struct sk_buff *skb); 4008 int dev_forward_skb_nomtu(struct net_device *dev, struct sk_buff *skb); 4009 bool is_skb_forwardable(const struct net_device *dev, 4010 const struct sk_buff *skb); 4011 4012 static __always_inline bool __is_skb_forwardable(const struct net_device *dev, 4013 const struct sk_buff *skb, 4014 const bool check_mtu) 4015 { 4016 const u32 vlan_hdr_len = 4; /* VLAN_HLEN */ 4017 unsigned int len; 4018 4019 if (!(dev->flags & IFF_UP)) 4020 return false; 4021 4022 if (!check_mtu) 4023 return true; 4024 4025 len = dev->mtu + dev->hard_header_len + vlan_hdr_len; 4026 if (skb->len <= len) 4027 return true; 4028 4029 /* if TSO is enabled, we don't care about the length as the packet 4030 * could be forwarded without being segmented before 4031 */ 4032 if (skb_is_gso(skb)) 4033 return true; 4034 4035 return false; 4036 } 4037 4038 static __always_inline int ____dev_forward_skb(struct net_device *dev, 4039 struct sk_buff *skb, 4040 const bool check_mtu) 4041 { 4042 if (skb_orphan_frags(skb, GFP_ATOMIC) || 4043 unlikely(!__is_skb_forwardable(dev, skb, check_mtu))) { 4044 atomic_long_inc(&dev->rx_dropped); 4045 kfree_skb(skb); 4046 return NET_RX_DROP; 4047 } 4048 4049 skb_scrub_packet(skb, true); 4050 skb->priority = 0; 4051 return 0; 4052 } 4053 4054 bool dev_nit_active(struct net_device *dev); 4055 void dev_queue_xmit_nit(struct sk_buff *skb, struct net_device *dev); 4056 4057 extern int netdev_budget; 4058 extern unsigned int netdev_budget_usecs; 4059 4060 /* Called by rtnetlink.c:rtnl_unlock() */ 4061 void netdev_run_todo(void); 4062 4063 /** 4064 * dev_put - release reference to device 4065 * @dev: network device 4066 * 4067 * Release reference to device to allow it to be freed. 4068 */ 4069 static inline void dev_put(struct net_device *dev) 4070 { 4071 #ifdef CONFIG_PCPU_DEV_REFCNT 4072 this_cpu_dec(*dev->pcpu_refcnt); 4073 #else 4074 refcount_dec(&dev->dev_refcnt); 4075 #endif 4076 } 4077 4078 /** 4079 * dev_hold - get reference to device 4080 * @dev: network device 4081 * 4082 * Hold reference to device to keep it from being freed. 4083 */ 4084 static inline void dev_hold(struct net_device *dev) 4085 { 4086 #ifdef CONFIG_PCPU_DEV_REFCNT 4087 this_cpu_inc(*dev->pcpu_refcnt); 4088 #else 4089 refcount_inc(&dev->dev_refcnt); 4090 #endif 4091 } 4092 4093 /* Carrier loss detection, dial on demand. The functions netif_carrier_on 4094 * and _off may be called from IRQ context, but it is caller 4095 * who is responsible for serialization of these calls. 4096 * 4097 * The name carrier is inappropriate, these functions should really be 4098 * called netif_lowerlayer_*() because they represent the state of any 4099 * kind of lower layer not just hardware media. 4100 */ 4101 4102 void linkwatch_init_dev(struct net_device *dev); 4103 void linkwatch_fire_event(struct net_device *dev); 4104 void linkwatch_forget_dev(struct net_device *dev); 4105 4106 /** 4107 * netif_carrier_ok - test if carrier present 4108 * @dev: network device 4109 * 4110 * Check if carrier is present on device 4111 */ 4112 static inline bool netif_carrier_ok(const struct net_device *dev) 4113 { 4114 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state); 4115 } 4116 4117 unsigned long dev_trans_start(struct net_device *dev); 4118 4119 void __netdev_watchdog_up(struct net_device *dev); 4120 4121 void netif_carrier_on(struct net_device *dev); 4122 4123 void netif_carrier_off(struct net_device *dev); 4124 4125 /** 4126 * netif_dormant_on - mark device as dormant. 4127 * @dev: network device 4128 * 4129 * Mark device as dormant (as per RFC2863). 4130 * 4131 * The dormant state indicates that the relevant interface is not 4132 * actually in a condition to pass packets (i.e., it is not 'up') but is 4133 * in a "pending" state, waiting for some external event. For "on- 4134 * demand" interfaces, this new state identifies the situation where the 4135 * interface is waiting for events to place it in the up state. 4136 */ 4137 static inline void netif_dormant_on(struct net_device *dev) 4138 { 4139 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state)) 4140 linkwatch_fire_event(dev); 4141 } 4142 4143 /** 4144 * netif_dormant_off - set device as not dormant. 4145 * @dev: network device 4146 * 4147 * Device is not in dormant state. 4148 */ 4149 static inline void netif_dormant_off(struct net_device *dev) 4150 { 4151 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state)) 4152 linkwatch_fire_event(dev); 4153 } 4154 4155 /** 4156 * netif_dormant - test if device is dormant 4157 * @dev: network device 4158 * 4159 * Check if device is dormant. 4160 */ 4161 static inline bool netif_dormant(const struct net_device *dev) 4162 { 4163 return test_bit(__LINK_STATE_DORMANT, &dev->state); 4164 } 4165 4166 4167 /** 4168 * netif_testing_on - mark device as under test. 4169 * @dev: network device 4170 * 4171 * Mark device as under test (as per RFC2863). 4172 * 4173 * The testing state indicates that some test(s) must be performed on 4174 * the interface. After completion, of the test, the interface state 4175 * will change to up, dormant, or down, as appropriate. 4176 */ 4177 static inline void netif_testing_on(struct net_device *dev) 4178 { 4179 if (!test_and_set_bit(__LINK_STATE_TESTING, &dev->state)) 4180 linkwatch_fire_event(dev); 4181 } 4182 4183 /** 4184 * netif_testing_off - set device as not under test. 4185 * @dev: network device 4186 * 4187 * Device is not in testing state. 4188 */ 4189 static inline void netif_testing_off(struct net_device *dev) 4190 { 4191 if (test_and_clear_bit(__LINK_STATE_TESTING, &dev->state)) 4192 linkwatch_fire_event(dev); 4193 } 4194 4195 /** 4196 * netif_testing - test if device is under test 4197 * @dev: network device 4198 * 4199 * Check if device is under test 4200 */ 4201 static inline bool netif_testing(const struct net_device *dev) 4202 { 4203 return test_bit(__LINK_STATE_TESTING, &dev->state); 4204 } 4205 4206 4207 /** 4208 * netif_oper_up - test if device is operational 4209 * @dev: network device 4210 * 4211 * Check if carrier is operational 4212 */ 4213 static inline bool netif_oper_up(const struct net_device *dev) 4214 { 4215 return (dev->operstate == IF_OPER_UP || 4216 dev->operstate == IF_OPER_UNKNOWN /* backward compat */); 4217 } 4218 4219 /** 4220 * netif_device_present - is device available or removed 4221 * @dev: network device 4222 * 4223 * Check if device has not been removed from system. 4224 */ 4225 static inline bool netif_device_present(const struct net_device *dev) 4226 { 4227 return test_bit(__LINK_STATE_PRESENT, &dev->state); 4228 } 4229 4230 void netif_device_detach(struct net_device *dev); 4231 4232 void netif_device_attach(struct net_device *dev); 4233 4234 /* 4235 * Network interface message level settings 4236 */ 4237 4238 enum { 4239 NETIF_MSG_DRV_BIT, 4240 NETIF_MSG_PROBE_BIT, 4241 NETIF_MSG_LINK_BIT, 4242 NETIF_MSG_TIMER_BIT, 4243 NETIF_MSG_IFDOWN_BIT, 4244 NETIF_MSG_IFUP_BIT, 4245 NETIF_MSG_RX_ERR_BIT, 4246 NETIF_MSG_TX_ERR_BIT, 4247 NETIF_MSG_TX_QUEUED_BIT, 4248 NETIF_MSG_INTR_BIT, 4249 NETIF_MSG_TX_DONE_BIT, 4250 NETIF_MSG_RX_STATUS_BIT, 4251 NETIF_MSG_PKTDATA_BIT, 4252 NETIF_MSG_HW_BIT, 4253 NETIF_MSG_WOL_BIT, 4254 4255 /* When you add a new bit above, update netif_msg_class_names array 4256 * in net/ethtool/common.c 4257 */ 4258 NETIF_MSG_CLASS_COUNT, 4259 }; 4260 /* Both ethtool_ops interface and internal driver implementation use u32 */ 4261 static_assert(NETIF_MSG_CLASS_COUNT <= 32); 4262 4263 #define __NETIF_MSG_BIT(bit) ((u32)1 << (bit)) 4264 #define __NETIF_MSG(name) __NETIF_MSG_BIT(NETIF_MSG_ ## name ## _BIT) 4265 4266 #define NETIF_MSG_DRV __NETIF_MSG(DRV) 4267 #define NETIF_MSG_PROBE __NETIF_MSG(PROBE) 4268 #define NETIF_MSG_LINK __NETIF_MSG(LINK) 4269 #define NETIF_MSG_TIMER __NETIF_MSG(TIMER) 4270 #define NETIF_MSG_IFDOWN __NETIF_MSG(IFDOWN) 4271 #define NETIF_MSG_IFUP __NETIF_MSG(IFUP) 4272 #define NETIF_MSG_RX_ERR __NETIF_MSG(RX_ERR) 4273 #define NETIF_MSG_TX_ERR __NETIF_MSG(TX_ERR) 4274 #define NETIF_MSG_TX_QUEUED __NETIF_MSG(TX_QUEUED) 4275 #define NETIF_MSG_INTR __NETIF_MSG(INTR) 4276 #define NETIF_MSG_TX_DONE __NETIF_MSG(TX_DONE) 4277 #define NETIF_MSG_RX_STATUS __NETIF_MSG(RX_STATUS) 4278 #define NETIF_MSG_PKTDATA __NETIF_MSG(PKTDATA) 4279 #define NETIF_MSG_HW __NETIF_MSG(HW) 4280 #define NETIF_MSG_WOL __NETIF_MSG(WOL) 4281 4282 #define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV) 4283 #define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE) 4284 #define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK) 4285 #define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER) 4286 #define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN) 4287 #define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP) 4288 #define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR) 4289 #define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR) 4290 #define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED) 4291 #define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR) 4292 #define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE) 4293 #define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS) 4294 #define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA) 4295 #define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW) 4296 #define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL) 4297 4298 static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits) 4299 { 4300 /* use default */ 4301 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8)) 4302 return default_msg_enable_bits; 4303 if (debug_value == 0) /* no output */ 4304 return 0; 4305 /* set low N bits */ 4306 return (1U << debug_value) - 1; 4307 } 4308 4309 static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu) 4310 { 4311 spin_lock(&txq->_xmit_lock); 4312 txq->xmit_lock_owner = cpu; 4313 } 4314 4315 static inline bool __netif_tx_acquire(struct netdev_queue *txq) 4316 { 4317 __acquire(&txq->_xmit_lock); 4318 return true; 4319 } 4320 4321 static inline void __netif_tx_release(struct netdev_queue *txq) 4322 { 4323 __release(&txq->_xmit_lock); 4324 } 4325 4326 static inline void __netif_tx_lock_bh(struct netdev_queue *txq) 4327 { 4328 spin_lock_bh(&txq->_xmit_lock); 4329 txq->xmit_lock_owner = smp_processor_id(); 4330 } 4331 4332 static inline bool __netif_tx_trylock(struct netdev_queue *txq) 4333 { 4334 bool ok = spin_trylock(&txq->_xmit_lock); 4335 if (likely(ok)) 4336 txq->xmit_lock_owner = smp_processor_id(); 4337 return ok; 4338 } 4339 4340 static inline void __netif_tx_unlock(struct netdev_queue *txq) 4341 { 4342 txq->xmit_lock_owner = -1; 4343 spin_unlock(&txq->_xmit_lock); 4344 } 4345 4346 static inline void __netif_tx_unlock_bh(struct netdev_queue *txq) 4347 { 4348 txq->xmit_lock_owner = -1; 4349 spin_unlock_bh(&txq->_xmit_lock); 4350 } 4351 4352 static inline void txq_trans_update(struct netdev_queue *txq) 4353 { 4354 if (txq->xmit_lock_owner != -1) 4355 txq->trans_start = jiffies; 4356 } 4357 4358 /* legacy drivers only, netdev_start_xmit() sets txq->trans_start */ 4359 static inline void netif_trans_update(struct net_device *dev) 4360 { 4361 struct netdev_queue *txq = netdev_get_tx_queue(dev, 0); 4362 4363 if (txq->trans_start != jiffies) 4364 txq->trans_start = jiffies; 4365 } 4366 4367 /** 4368 * netif_tx_lock - grab network device transmit lock 4369 * @dev: network device 4370 * 4371 * Get network device transmit lock 4372 */ 4373 static inline void netif_tx_lock(struct net_device *dev) 4374 { 4375 unsigned int i; 4376 int cpu; 4377 4378 spin_lock(&dev->tx_global_lock); 4379 cpu = smp_processor_id(); 4380 for (i = 0; i < dev->num_tx_queues; i++) { 4381 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 4382 4383 /* We are the only thread of execution doing a 4384 * freeze, but we have to grab the _xmit_lock in 4385 * order to synchronize with threads which are in 4386 * the ->hard_start_xmit() handler and already 4387 * checked the frozen bit. 4388 */ 4389 __netif_tx_lock(txq, cpu); 4390 set_bit(__QUEUE_STATE_FROZEN, &txq->state); 4391 __netif_tx_unlock(txq); 4392 } 4393 } 4394 4395 static inline void netif_tx_lock_bh(struct net_device *dev) 4396 { 4397 local_bh_disable(); 4398 netif_tx_lock(dev); 4399 } 4400 4401 static inline void netif_tx_unlock(struct net_device *dev) 4402 { 4403 unsigned int i; 4404 4405 for (i = 0; i < dev->num_tx_queues; i++) { 4406 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 4407 4408 /* No need to grab the _xmit_lock here. If the 4409 * queue is not stopped for another reason, we 4410 * force a schedule. 4411 */ 4412 clear_bit(__QUEUE_STATE_FROZEN, &txq->state); 4413 netif_schedule_queue(txq); 4414 } 4415 spin_unlock(&dev->tx_global_lock); 4416 } 4417 4418 static inline void netif_tx_unlock_bh(struct net_device *dev) 4419 { 4420 netif_tx_unlock(dev); 4421 local_bh_enable(); 4422 } 4423 4424 #define HARD_TX_LOCK(dev, txq, cpu) { \ 4425 if ((dev->features & NETIF_F_LLTX) == 0) { \ 4426 __netif_tx_lock(txq, cpu); \ 4427 } else { \ 4428 __netif_tx_acquire(txq); \ 4429 } \ 4430 } 4431 4432 #define HARD_TX_TRYLOCK(dev, txq) \ 4433 (((dev->features & NETIF_F_LLTX) == 0) ? \ 4434 __netif_tx_trylock(txq) : \ 4435 __netif_tx_acquire(txq)) 4436 4437 #define HARD_TX_UNLOCK(dev, txq) { \ 4438 if ((dev->features & NETIF_F_LLTX) == 0) { \ 4439 __netif_tx_unlock(txq); \ 4440 } else { \ 4441 __netif_tx_release(txq); \ 4442 } \ 4443 } 4444 4445 static inline void netif_tx_disable(struct net_device *dev) 4446 { 4447 unsigned int i; 4448 int cpu; 4449 4450 local_bh_disable(); 4451 cpu = smp_processor_id(); 4452 spin_lock(&dev->tx_global_lock); 4453 for (i = 0; i < dev->num_tx_queues; i++) { 4454 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 4455 4456 __netif_tx_lock(txq, cpu); 4457 netif_tx_stop_queue(txq); 4458 __netif_tx_unlock(txq); 4459 } 4460 spin_unlock(&dev->tx_global_lock); 4461 local_bh_enable(); 4462 } 4463 4464 static inline void netif_addr_lock(struct net_device *dev) 4465 { 4466 unsigned char nest_level = 0; 4467 4468 #ifdef CONFIG_LOCKDEP 4469 nest_level = dev->nested_level; 4470 #endif 4471 spin_lock_nested(&dev->addr_list_lock, nest_level); 4472 } 4473 4474 static inline void netif_addr_lock_bh(struct net_device *dev) 4475 { 4476 unsigned char nest_level = 0; 4477 4478 #ifdef CONFIG_LOCKDEP 4479 nest_level = dev->nested_level; 4480 #endif 4481 local_bh_disable(); 4482 spin_lock_nested(&dev->addr_list_lock, nest_level); 4483 } 4484 4485 static inline void netif_addr_unlock(struct net_device *dev) 4486 { 4487 spin_unlock(&dev->addr_list_lock); 4488 } 4489 4490 static inline void netif_addr_unlock_bh(struct net_device *dev) 4491 { 4492 spin_unlock_bh(&dev->addr_list_lock); 4493 } 4494 4495 /* 4496 * dev_addrs walker. Should be used only for read access. Call with 4497 * rcu_read_lock held. 4498 */ 4499 #define for_each_dev_addr(dev, ha) \ 4500 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list) 4501 4502 /* These functions live elsewhere (drivers/net/net_init.c, but related) */ 4503 4504 void ether_setup(struct net_device *dev); 4505 4506 /* Support for loadable net-drivers */ 4507 struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name, 4508 unsigned char name_assign_type, 4509 void (*setup)(struct net_device *), 4510 unsigned int txqs, unsigned int rxqs); 4511 #define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \ 4512 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1) 4513 4514 #define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \ 4515 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \ 4516 count) 4517 4518 int register_netdev(struct net_device *dev); 4519 void unregister_netdev(struct net_device *dev); 4520 4521 int devm_register_netdev(struct device *dev, struct net_device *ndev); 4522 4523 /* General hardware address lists handling functions */ 4524 int __hw_addr_sync(struct netdev_hw_addr_list *to_list, 4525 struct netdev_hw_addr_list *from_list, int addr_len); 4526 void __hw_addr_unsync(struct netdev_hw_addr_list *to_list, 4527 struct netdev_hw_addr_list *from_list, int addr_len); 4528 int __hw_addr_sync_dev(struct netdev_hw_addr_list *list, 4529 struct net_device *dev, 4530 int (*sync)(struct net_device *, const unsigned char *), 4531 int (*unsync)(struct net_device *, 4532 const unsigned char *)); 4533 int __hw_addr_ref_sync_dev(struct netdev_hw_addr_list *list, 4534 struct net_device *dev, 4535 int (*sync)(struct net_device *, 4536 const unsigned char *, int), 4537 int (*unsync)(struct net_device *, 4538 const unsigned char *, int)); 4539 void __hw_addr_ref_unsync_dev(struct netdev_hw_addr_list *list, 4540 struct net_device *dev, 4541 int (*unsync)(struct net_device *, 4542 const unsigned char *, int)); 4543 void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list, 4544 struct net_device *dev, 4545 int (*unsync)(struct net_device *, 4546 const unsigned char *)); 4547 void __hw_addr_init(struct netdev_hw_addr_list *list); 4548 4549 /* Functions used for device addresses handling */ 4550 int dev_addr_add(struct net_device *dev, const unsigned char *addr, 4551 unsigned char addr_type); 4552 int dev_addr_del(struct net_device *dev, const unsigned char *addr, 4553 unsigned char addr_type); 4554 void dev_addr_flush(struct net_device *dev); 4555 int dev_addr_init(struct net_device *dev); 4556 4557 /* Functions used for unicast addresses handling */ 4558 int dev_uc_add(struct net_device *dev, const unsigned char *addr); 4559 int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr); 4560 int dev_uc_del(struct net_device *dev, const unsigned char *addr); 4561 int dev_uc_sync(struct net_device *to, struct net_device *from); 4562 int dev_uc_sync_multiple(struct net_device *to, struct net_device *from); 4563 void dev_uc_unsync(struct net_device *to, struct net_device *from); 4564 void dev_uc_flush(struct net_device *dev); 4565 void dev_uc_init(struct net_device *dev); 4566 4567 /** 4568 * __dev_uc_sync - Synchonize device's unicast list 4569 * @dev: device to sync 4570 * @sync: function to call if address should be added 4571 * @unsync: function to call if address should be removed 4572 * 4573 * Add newly added addresses to the interface, and release 4574 * addresses that have been deleted. 4575 */ 4576 static inline int __dev_uc_sync(struct net_device *dev, 4577 int (*sync)(struct net_device *, 4578 const unsigned char *), 4579 int (*unsync)(struct net_device *, 4580 const unsigned char *)) 4581 { 4582 return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync); 4583 } 4584 4585 /** 4586 * __dev_uc_unsync - Remove synchronized addresses from device 4587 * @dev: device to sync 4588 * @unsync: function to call if address should be removed 4589 * 4590 * Remove all addresses that were added to the device by dev_uc_sync(). 4591 */ 4592 static inline void __dev_uc_unsync(struct net_device *dev, 4593 int (*unsync)(struct net_device *, 4594 const unsigned char *)) 4595 { 4596 __hw_addr_unsync_dev(&dev->uc, dev, unsync); 4597 } 4598 4599 /* Functions used for multicast addresses handling */ 4600 int dev_mc_add(struct net_device *dev, const unsigned char *addr); 4601 int dev_mc_add_global(struct net_device *dev, const unsigned char *addr); 4602 int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr); 4603 int dev_mc_del(struct net_device *dev, const unsigned char *addr); 4604 int dev_mc_del_global(struct net_device *dev, const unsigned char *addr); 4605 int dev_mc_sync(struct net_device *to, struct net_device *from); 4606 int dev_mc_sync_multiple(struct net_device *to, struct net_device *from); 4607 void dev_mc_unsync(struct net_device *to, struct net_device *from); 4608 void dev_mc_flush(struct net_device *dev); 4609 void dev_mc_init(struct net_device *dev); 4610 4611 /** 4612 * __dev_mc_sync - Synchonize device's multicast list 4613 * @dev: device to sync 4614 * @sync: function to call if address should be added 4615 * @unsync: function to call if address should be removed 4616 * 4617 * Add newly added addresses to the interface, and release 4618 * addresses that have been deleted. 4619 */ 4620 static inline int __dev_mc_sync(struct net_device *dev, 4621 int (*sync)(struct net_device *, 4622 const unsigned char *), 4623 int (*unsync)(struct net_device *, 4624 const unsigned char *)) 4625 { 4626 return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync); 4627 } 4628 4629 /** 4630 * __dev_mc_unsync - Remove synchronized addresses from device 4631 * @dev: device to sync 4632 * @unsync: function to call if address should be removed 4633 * 4634 * Remove all addresses that were added to the device by dev_mc_sync(). 4635 */ 4636 static inline void __dev_mc_unsync(struct net_device *dev, 4637 int (*unsync)(struct net_device *, 4638 const unsigned char *)) 4639 { 4640 __hw_addr_unsync_dev(&dev->mc, dev, unsync); 4641 } 4642 4643 /* Functions used for secondary unicast and multicast support */ 4644 void dev_set_rx_mode(struct net_device *dev); 4645 void __dev_set_rx_mode(struct net_device *dev); 4646 int dev_set_promiscuity(struct net_device *dev, int inc); 4647 int dev_set_allmulti(struct net_device *dev, int inc); 4648 void netdev_state_change(struct net_device *dev); 4649 void __netdev_notify_peers(struct net_device *dev); 4650 void netdev_notify_peers(struct net_device *dev); 4651 void netdev_features_change(struct net_device *dev); 4652 /* Load a device via the kmod */ 4653 void dev_load(struct net *net, const char *name); 4654 struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev, 4655 struct rtnl_link_stats64 *storage); 4656 void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64, 4657 const struct net_device_stats *netdev_stats); 4658 void dev_fetch_sw_netstats(struct rtnl_link_stats64 *s, 4659 const struct pcpu_sw_netstats __percpu *netstats); 4660 void dev_get_tstats64(struct net_device *dev, struct rtnl_link_stats64 *s); 4661 4662 extern int netdev_max_backlog; 4663 extern int netdev_tstamp_prequeue; 4664 extern int weight_p; 4665 extern int dev_weight_rx_bias; 4666 extern int dev_weight_tx_bias; 4667 extern int dev_rx_weight; 4668 extern int dev_tx_weight; 4669 extern int gro_normal_batch; 4670 4671 enum { 4672 NESTED_SYNC_IMM_BIT, 4673 NESTED_SYNC_TODO_BIT, 4674 }; 4675 4676 #define __NESTED_SYNC_BIT(bit) ((u32)1 << (bit)) 4677 #define __NESTED_SYNC(name) __NESTED_SYNC_BIT(NESTED_SYNC_ ## name ## _BIT) 4678 4679 #define NESTED_SYNC_IMM __NESTED_SYNC(IMM) 4680 #define NESTED_SYNC_TODO __NESTED_SYNC(TODO) 4681 4682 struct netdev_nested_priv { 4683 unsigned char flags; 4684 void *data; 4685 }; 4686 4687 bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev); 4688 struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev, 4689 struct list_head **iter); 4690 struct net_device *netdev_all_upper_get_next_dev_rcu(struct net_device *dev, 4691 struct list_head **iter); 4692 4693 #ifdef CONFIG_LOCKDEP 4694 static LIST_HEAD(net_unlink_list); 4695 4696 static inline void net_unlink_todo(struct net_device *dev) 4697 { 4698 if (list_empty(&dev->unlink_list)) 4699 list_add_tail(&dev->unlink_list, &net_unlink_list); 4700 } 4701 #endif 4702 4703 /* iterate through upper list, must be called under RCU read lock */ 4704 #define netdev_for_each_upper_dev_rcu(dev, updev, iter) \ 4705 for (iter = &(dev)->adj_list.upper, \ 4706 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \ 4707 updev; \ 4708 updev = netdev_upper_get_next_dev_rcu(dev, &(iter))) 4709 4710 int netdev_walk_all_upper_dev_rcu(struct net_device *dev, 4711 int (*fn)(struct net_device *upper_dev, 4712 struct netdev_nested_priv *priv), 4713 struct netdev_nested_priv *priv); 4714 4715 bool netdev_has_upper_dev_all_rcu(struct net_device *dev, 4716 struct net_device *upper_dev); 4717 4718 bool netdev_has_any_upper_dev(struct net_device *dev); 4719 4720 void *netdev_lower_get_next_private(struct net_device *dev, 4721 struct list_head **iter); 4722 void *netdev_lower_get_next_private_rcu(struct net_device *dev, 4723 struct list_head **iter); 4724 4725 #define netdev_for_each_lower_private(dev, priv, iter) \ 4726 for (iter = (dev)->adj_list.lower.next, \ 4727 priv = netdev_lower_get_next_private(dev, &(iter)); \ 4728 priv; \ 4729 priv = netdev_lower_get_next_private(dev, &(iter))) 4730 4731 #define netdev_for_each_lower_private_rcu(dev, priv, iter) \ 4732 for (iter = &(dev)->adj_list.lower, \ 4733 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \ 4734 priv; \ 4735 priv = netdev_lower_get_next_private_rcu(dev, &(iter))) 4736 4737 void *netdev_lower_get_next(struct net_device *dev, 4738 struct list_head **iter); 4739 4740 #define netdev_for_each_lower_dev(dev, ldev, iter) \ 4741 for (iter = (dev)->adj_list.lower.next, \ 4742 ldev = netdev_lower_get_next(dev, &(iter)); \ 4743 ldev; \ 4744 ldev = netdev_lower_get_next(dev, &(iter))) 4745 4746 struct net_device *netdev_next_lower_dev_rcu(struct net_device *dev, 4747 struct list_head **iter); 4748 int netdev_walk_all_lower_dev(struct net_device *dev, 4749 int (*fn)(struct net_device *lower_dev, 4750 struct netdev_nested_priv *priv), 4751 struct netdev_nested_priv *priv); 4752 int netdev_walk_all_lower_dev_rcu(struct net_device *dev, 4753 int (*fn)(struct net_device *lower_dev, 4754 struct netdev_nested_priv *priv), 4755 struct netdev_nested_priv *priv); 4756 4757 void *netdev_adjacent_get_private(struct list_head *adj_list); 4758 void *netdev_lower_get_first_private_rcu(struct net_device *dev); 4759 struct net_device *netdev_master_upper_dev_get(struct net_device *dev); 4760 struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev); 4761 int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev, 4762 struct netlink_ext_ack *extack); 4763 int netdev_master_upper_dev_link(struct net_device *dev, 4764 struct net_device *upper_dev, 4765 void *upper_priv, void *upper_info, 4766 struct netlink_ext_ack *extack); 4767 void netdev_upper_dev_unlink(struct net_device *dev, 4768 struct net_device *upper_dev); 4769 int netdev_adjacent_change_prepare(struct net_device *old_dev, 4770 struct net_device *new_dev, 4771 struct net_device *dev, 4772 struct netlink_ext_ack *extack); 4773 void netdev_adjacent_change_commit(struct net_device *old_dev, 4774 struct net_device *new_dev, 4775 struct net_device *dev); 4776 void netdev_adjacent_change_abort(struct net_device *old_dev, 4777 struct net_device *new_dev, 4778 struct net_device *dev); 4779 void netdev_adjacent_rename_links(struct net_device *dev, char *oldname); 4780 void *netdev_lower_dev_get_private(struct net_device *dev, 4781 struct net_device *lower_dev); 4782 void netdev_lower_state_changed(struct net_device *lower_dev, 4783 void *lower_state_info); 4784 4785 /* RSS keys are 40 or 52 bytes long */ 4786 #define NETDEV_RSS_KEY_LEN 52 4787 extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN] __read_mostly; 4788 void netdev_rss_key_fill(void *buffer, size_t len); 4789 4790 int skb_checksum_help(struct sk_buff *skb); 4791 int skb_crc32c_csum_help(struct sk_buff *skb); 4792 int skb_csum_hwoffload_help(struct sk_buff *skb, 4793 const netdev_features_t features); 4794 4795 struct sk_buff *__skb_gso_segment(struct sk_buff *skb, 4796 netdev_features_t features, bool tx_path); 4797 struct sk_buff *skb_mac_gso_segment(struct sk_buff *skb, 4798 netdev_features_t features); 4799 4800 struct netdev_bonding_info { 4801 ifslave slave; 4802 ifbond master; 4803 }; 4804 4805 struct netdev_notifier_bonding_info { 4806 struct netdev_notifier_info info; /* must be first */ 4807 struct netdev_bonding_info bonding_info; 4808 }; 4809 4810 void netdev_bonding_info_change(struct net_device *dev, 4811 struct netdev_bonding_info *bonding_info); 4812 4813 #if IS_ENABLED(CONFIG_ETHTOOL_NETLINK) 4814 void ethtool_notify(struct net_device *dev, unsigned int cmd, const void *data); 4815 #else 4816 static inline void ethtool_notify(struct net_device *dev, unsigned int cmd, 4817 const void *data) 4818 { 4819 } 4820 #endif 4821 4822 static inline 4823 struct sk_buff *skb_gso_segment(struct sk_buff *skb, netdev_features_t features) 4824 { 4825 return __skb_gso_segment(skb, features, true); 4826 } 4827 __be16 skb_network_protocol(struct sk_buff *skb, int *depth); 4828 4829 static inline bool can_checksum_protocol(netdev_features_t features, 4830 __be16 protocol) 4831 { 4832 if (protocol == htons(ETH_P_FCOE)) 4833 return !!(features & NETIF_F_FCOE_CRC); 4834 4835 /* Assume this is an IP checksum (not SCTP CRC) */ 4836 4837 if (features & NETIF_F_HW_CSUM) { 4838 /* Can checksum everything */ 4839 return true; 4840 } 4841 4842 switch (protocol) { 4843 case htons(ETH_P_IP): 4844 return !!(features & NETIF_F_IP_CSUM); 4845 case htons(ETH_P_IPV6): 4846 return !!(features & NETIF_F_IPV6_CSUM); 4847 default: 4848 return false; 4849 } 4850 } 4851 4852 #ifdef CONFIG_BUG 4853 void netdev_rx_csum_fault(struct net_device *dev, struct sk_buff *skb); 4854 #else 4855 static inline void netdev_rx_csum_fault(struct net_device *dev, 4856 struct sk_buff *skb) 4857 { 4858 } 4859 #endif 4860 /* rx skb timestamps */ 4861 void net_enable_timestamp(void); 4862 void net_disable_timestamp(void); 4863 4864 #ifdef CONFIG_PROC_FS 4865 int __init dev_proc_init(void); 4866 #else 4867 #define dev_proc_init() 0 4868 #endif 4869 4870 static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops, 4871 struct sk_buff *skb, struct net_device *dev, 4872 bool more) 4873 { 4874 __this_cpu_write(softnet_data.xmit.more, more); 4875 return ops->ndo_start_xmit(skb, dev); 4876 } 4877 4878 static inline bool netdev_xmit_more(void) 4879 { 4880 return __this_cpu_read(softnet_data.xmit.more); 4881 } 4882 4883 static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev, 4884 struct netdev_queue *txq, bool more) 4885 { 4886 const struct net_device_ops *ops = dev->netdev_ops; 4887 netdev_tx_t rc; 4888 4889 rc = __netdev_start_xmit(ops, skb, dev, more); 4890 if (rc == NETDEV_TX_OK) 4891 txq_trans_update(txq); 4892 4893 return rc; 4894 } 4895 4896 int netdev_class_create_file_ns(const struct class_attribute *class_attr, 4897 const void *ns); 4898 void netdev_class_remove_file_ns(const struct class_attribute *class_attr, 4899 const void *ns); 4900 4901 extern const struct kobj_ns_type_operations net_ns_type_operations; 4902 4903 const char *netdev_drivername(const struct net_device *dev); 4904 4905 void linkwatch_run_queue(void); 4906 4907 static inline netdev_features_t netdev_intersect_features(netdev_features_t f1, 4908 netdev_features_t f2) 4909 { 4910 if ((f1 ^ f2) & NETIF_F_HW_CSUM) { 4911 if (f1 & NETIF_F_HW_CSUM) 4912 f1 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM); 4913 else 4914 f2 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM); 4915 } 4916 4917 return f1 & f2; 4918 } 4919 4920 static inline netdev_features_t netdev_get_wanted_features( 4921 struct net_device *dev) 4922 { 4923 return (dev->features & ~dev->hw_features) | dev->wanted_features; 4924 } 4925 netdev_features_t netdev_increment_features(netdev_features_t all, 4926 netdev_features_t one, netdev_features_t mask); 4927 4928 /* Allow TSO being used on stacked device : 4929 * Performing the GSO segmentation before last device 4930 * is a performance improvement. 4931 */ 4932 static inline netdev_features_t netdev_add_tso_features(netdev_features_t features, 4933 netdev_features_t mask) 4934 { 4935 return netdev_increment_features(features, NETIF_F_ALL_TSO, mask); 4936 } 4937 4938 int __netdev_update_features(struct net_device *dev); 4939 void netdev_update_features(struct net_device *dev); 4940 void netdev_change_features(struct net_device *dev); 4941 4942 void netif_stacked_transfer_operstate(const struct net_device *rootdev, 4943 struct net_device *dev); 4944 4945 netdev_features_t passthru_features_check(struct sk_buff *skb, 4946 struct net_device *dev, 4947 netdev_features_t features); 4948 netdev_features_t netif_skb_features(struct sk_buff *skb); 4949 4950 static inline bool net_gso_ok(netdev_features_t features, int gso_type) 4951 { 4952 netdev_features_t feature = (netdev_features_t)gso_type << NETIF_F_GSO_SHIFT; 4953 4954 /* check flags correspondence */ 4955 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT)); 4956 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT)); 4957 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT)); 4958 BUILD_BUG_ON(SKB_GSO_TCP_FIXEDID != (NETIF_F_TSO_MANGLEID >> NETIF_F_GSO_SHIFT)); 4959 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT)); 4960 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT)); 4961 BUILD_BUG_ON(SKB_GSO_GRE != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT)); 4962 BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT)); 4963 BUILD_BUG_ON(SKB_GSO_IPXIP4 != (NETIF_F_GSO_IPXIP4 >> NETIF_F_GSO_SHIFT)); 4964 BUILD_BUG_ON(SKB_GSO_IPXIP6 != (NETIF_F_GSO_IPXIP6 >> NETIF_F_GSO_SHIFT)); 4965 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT)); 4966 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT)); 4967 BUILD_BUG_ON(SKB_GSO_PARTIAL != (NETIF_F_GSO_PARTIAL >> NETIF_F_GSO_SHIFT)); 4968 BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT)); 4969 BUILD_BUG_ON(SKB_GSO_SCTP != (NETIF_F_GSO_SCTP >> NETIF_F_GSO_SHIFT)); 4970 BUILD_BUG_ON(SKB_GSO_ESP != (NETIF_F_GSO_ESP >> NETIF_F_GSO_SHIFT)); 4971 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_GSO_UDP >> NETIF_F_GSO_SHIFT)); 4972 BUILD_BUG_ON(SKB_GSO_UDP_L4 != (NETIF_F_GSO_UDP_L4 >> NETIF_F_GSO_SHIFT)); 4973 BUILD_BUG_ON(SKB_GSO_FRAGLIST != (NETIF_F_GSO_FRAGLIST >> NETIF_F_GSO_SHIFT)); 4974 4975 return (features & feature) == feature; 4976 } 4977 4978 static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features) 4979 { 4980 return net_gso_ok(features, skb_shinfo(skb)->gso_type) && 4981 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST)); 4982 } 4983 4984 static inline bool netif_needs_gso(struct sk_buff *skb, 4985 netdev_features_t features) 4986 { 4987 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) || 4988 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) && 4989 (skb->ip_summed != CHECKSUM_UNNECESSARY))); 4990 } 4991 4992 static inline void netif_set_gso_max_size(struct net_device *dev, 4993 unsigned int size) 4994 { 4995 dev->gso_max_size = size; 4996 } 4997 4998 static inline void skb_gso_error_unwind(struct sk_buff *skb, __be16 protocol, 4999 int pulled_hlen, u16 mac_offset, 5000 int mac_len) 5001 { 5002 skb->protocol = protocol; 5003 skb->encapsulation = 1; 5004 skb_push(skb, pulled_hlen); 5005 skb_reset_transport_header(skb); 5006 skb->mac_header = mac_offset; 5007 skb->network_header = skb->mac_header + mac_len; 5008 skb->mac_len = mac_len; 5009 } 5010 5011 static inline bool netif_is_macsec(const struct net_device *dev) 5012 { 5013 return dev->priv_flags & IFF_MACSEC; 5014 } 5015 5016 static inline bool netif_is_macvlan(const struct net_device *dev) 5017 { 5018 return dev->priv_flags & IFF_MACVLAN; 5019 } 5020 5021 static inline bool netif_is_macvlan_port(const struct net_device *dev) 5022 { 5023 return dev->priv_flags & IFF_MACVLAN_PORT; 5024 } 5025 5026 static inline bool netif_is_bond_master(const struct net_device *dev) 5027 { 5028 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING; 5029 } 5030 5031 static inline bool netif_is_bond_slave(const struct net_device *dev) 5032 { 5033 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING; 5034 } 5035 5036 static inline bool netif_supports_nofcs(struct net_device *dev) 5037 { 5038 return dev->priv_flags & IFF_SUPP_NOFCS; 5039 } 5040 5041 static inline bool netif_has_l3_rx_handler(const struct net_device *dev) 5042 { 5043 return dev->priv_flags & IFF_L3MDEV_RX_HANDLER; 5044 } 5045 5046 static inline bool netif_is_l3_master(const struct net_device *dev) 5047 { 5048 return dev->priv_flags & IFF_L3MDEV_MASTER; 5049 } 5050 5051 static inline bool netif_is_l3_slave(const struct net_device *dev) 5052 { 5053 return dev->priv_flags & IFF_L3MDEV_SLAVE; 5054 } 5055 5056 static inline bool netif_is_bridge_master(const struct net_device *dev) 5057 { 5058 return dev->priv_flags & IFF_EBRIDGE; 5059 } 5060 5061 static inline bool netif_is_bridge_port(const struct net_device *dev) 5062 { 5063 return dev->priv_flags & IFF_BRIDGE_PORT; 5064 } 5065 5066 static inline bool netif_is_ovs_master(const struct net_device *dev) 5067 { 5068 return dev->priv_flags & IFF_OPENVSWITCH; 5069 } 5070 5071 static inline bool netif_is_ovs_port(const struct net_device *dev) 5072 { 5073 return dev->priv_flags & IFF_OVS_DATAPATH; 5074 } 5075 5076 static inline bool netif_is_any_bridge_port(const struct net_device *dev) 5077 { 5078 return netif_is_bridge_port(dev) || netif_is_ovs_port(dev); 5079 } 5080 5081 static inline bool netif_is_team_master(const struct net_device *dev) 5082 { 5083 return dev->priv_flags & IFF_TEAM; 5084 } 5085 5086 static inline bool netif_is_team_port(const struct net_device *dev) 5087 { 5088 return dev->priv_flags & IFF_TEAM_PORT; 5089 } 5090 5091 static inline bool netif_is_lag_master(const struct net_device *dev) 5092 { 5093 return netif_is_bond_master(dev) || netif_is_team_master(dev); 5094 } 5095 5096 static inline bool netif_is_lag_port(const struct net_device *dev) 5097 { 5098 return netif_is_bond_slave(dev) || netif_is_team_port(dev); 5099 } 5100 5101 static inline bool netif_is_rxfh_configured(const struct net_device *dev) 5102 { 5103 return dev->priv_flags & IFF_RXFH_CONFIGURED; 5104 } 5105 5106 static inline bool netif_is_failover(const struct net_device *dev) 5107 { 5108 return dev->priv_flags & IFF_FAILOVER; 5109 } 5110 5111 static inline bool netif_is_failover_slave(const struct net_device *dev) 5112 { 5113 return dev->priv_flags & IFF_FAILOVER_SLAVE; 5114 } 5115 5116 /* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */ 5117 static inline void netif_keep_dst(struct net_device *dev) 5118 { 5119 dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM); 5120 } 5121 5122 /* return true if dev can't cope with mtu frames that need vlan tag insertion */ 5123 static inline bool netif_reduces_vlan_mtu(struct net_device *dev) 5124 { 5125 /* TODO: reserve and use an additional IFF bit, if we get more users */ 5126 return dev->priv_flags & IFF_MACSEC; 5127 } 5128 5129 extern struct pernet_operations __net_initdata loopback_net_ops; 5130 5131 /* Logging, debugging and troubleshooting/diagnostic helpers. */ 5132 5133 /* netdev_printk helpers, similar to dev_printk */ 5134 5135 static inline const char *netdev_name(const struct net_device *dev) 5136 { 5137 if (!dev->name[0] || strchr(dev->name, '%')) 5138 return "(unnamed net_device)"; 5139 return dev->name; 5140 } 5141 5142 static inline bool netdev_unregistering(const struct net_device *dev) 5143 { 5144 return dev->reg_state == NETREG_UNREGISTERING; 5145 } 5146 5147 static inline const char *netdev_reg_state(const struct net_device *dev) 5148 { 5149 switch (dev->reg_state) { 5150 case NETREG_UNINITIALIZED: return " (uninitialized)"; 5151 case NETREG_REGISTERED: return ""; 5152 case NETREG_UNREGISTERING: return " (unregistering)"; 5153 case NETREG_UNREGISTERED: return " (unregistered)"; 5154 case NETREG_RELEASED: return " (released)"; 5155 case NETREG_DUMMY: return " (dummy)"; 5156 } 5157 5158 WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, dev->reg_state); 5159 return " (unknown)"; 5160 } 5161 5162 __printf(3, 4) __cold 5163 void netdev_printk(const char *level, const struct net_device *dev, 5164 const char *format, ...); 5165 __printf(2, 3) __cold 5166 void netdev_emerg(const struct net_device *dev, const char *format, ...); 5167 __printf(2, 3) __cold 5168 void netdev_alert(const struct net_device *dev, const char *format, ...); 5169 __printf(2, 3) __cold 5170 void netdev_crit(const struct net_device *dev, const char *format, ...); 5171 __printf(2, 3) __cold 5172 void netdev_err(const struct net_device *dev, const char *format, ...); 5173 __printf(2, 3) __cold 5174 void netdev_warn(const struct net_device *dev, const char *format, ...); 5175 __printf(2, 3) __cold 5176 void netdev_notice(const struct net_device *dev, const char *format, ...); 5177 __printf(2, 3) __cold 5178 void netdev_info(const struct net_device *dev, const char *format, ...); 5179 5180 #define netdev_level_once(level, dev, fmt, ...) \ 5181 do { \ 5182 static bool __print_once __read_mostly; \ 5183 \ 5184 if (!__print_once) { \ 5185 __print_once = true; \ 5186 netdev_printk(level, dev, fmt, ##__VA_ARGS__); \ 5187 } \ 5188 } while (0) 5189 5190 #define netdev_emerg_once(dev, fmt, ...) \ 5191 netdev_level_once(KERN_EMERG, dev, fmt, ##__VA_ARGS__) 5192 #define netdev_alert_once(dev, fmt, ...) \ 5193 netdev_level_once(KERN_ALERT, dev, fmt, ##__VA_ARGS__) 5194 #define netdev_crit_once(dev, fmt, ...) \ 5195 netdev_level_once(KERN_CRIT, dev, fmt, ##__VA_ARGS__) 5196 #define netdev_err_once(dev, fmt, ...) \ 5197 netdev_level_once(KERN_ERR, dev, fmt, ##__VA_ARGS__) 5198 #define netdev_warn_once(dev, fmt, ...) \ 5199 netdev_level_once(KERN_WARNING, dev, fmt, ##__VA_ARGS__) 5200 #define netdev_notice_once(dev, fmt, ...) \ 5201 netdev_level_once(KERN_NOTICE, dev, fmt, ##__VA_ARGS__) 5202 #define netdev_info_once(dev, fmt, ...) \ 5203 netdev_level_once(KERN_INFO, dev, fmt, ##__VA_ARGS__) 5204 5205 #define MODULE_ALIAS_NETDEV(device) \ 5206 MODULE_ALIAS("netdev-" device) 5207 5208 #if defined(CONFIG_DYNAMIC_DEBUG) || \ 5209 (defined(CONFIG_DYNAMIC_DEBUG_CORE) && defined(DYNAMIC_DEBUG_MODULE)) 5210 #define netdev_dbg(__dev, format, args...) \ 5211 do { \ 5212 dynamic_netdev_dbg(__dev, format, ##args); \ 5213 } while (0) 5214 #elif defined(DEBUG) 5215 #define netdev_dbg(__dev, format, args...) \ 5216 netdev_printk(KERN_DEBUG, __dev, format, ##args) 5217 #else 5218 #define netdev_dbg(__dev, format, args...) \ 5219 ({ \ 5220 if (0) \ 5221 netdev_printk(KERN_DEBUG, __dev, format, ##args); \ 5222 }) 5223 #endif 5224 5225 #if defined(VERBOSE_DEBUG) 5226 #define netdev_vdbg netdev_dbg 5227 #else 5228 5229 #define netdev_vdbg(dev, format, args...) \ 5230 ({ \ 5231 if (0) \ 5232 netdev_printk(KERN_DEBUG, dev, format, ##args); \ 5233 0; \ 5234 }) 5235 #endif 5236 5237 /* 5238 * netdev_WARN() acts like dev_printk(), but with the key difference 5239 * of using a WARN/WARN_ON to get the message out, including the 5240 * file/line information and a backtrace. 5241 */ 5242 #define netdev_WARN(dev, format, args...) \ 5243 WARN(1, "netdevice: %s%s: " format, netdev_name(dev), \ 5244 netdev_reg_state(dev), ##args) 5245 5246 #define netdev_WARN_ONCE(dev, format, args...) \ 5247 WARN_ONCE(1, "netdevice: %s%s: " format, netdev_name(dev), \ 5248 netdev_reg_state(dev), ##args) 5249 5250 /* netif printk helpers, similar to netdev_printk */ 5251 5252 #define netif_printk(priv, type, level, dev, fmt, args...) \ 5253 do { \ 5254 if (netif_msg_##type(priv)) \ 5255 netdev_printk(level, (dev), fmt, ##args); \ 5256 } while (0) 5257 5258 #define netif_level(level, priv, type, dev, fmt, args...) \ 5259 do { \ 5260 if (netif_msg_##type(priv)) \ 5261 netdev_##level(dev, fmt, ##args); \ 5262 } while (0) 5263 5264 #define netif_emerg(priv, type, dev, fmt, args...) \ 5265 netif_level(emerg, priv, type, dev, fmt, ##args) 5266 #define netif_alert(priv, type, dev, fmt, args...) \ 5267 netif_level(alert, priv, type, dev, fmt, ##args) 5268 #define netif_crit(priv, type, dev, fmt, args...) \ 5269 netif_level(crit, priv, type, dev, fmt, ##args) 5270 #define netif_err(priv, type, dev, fmt, args...) \ 5271 netif_level(err, priv, type, dev, fmt, ##args) 5272 #define netif_warn(priv, type, dev, fmt, args...) \ 5273 netif_level(warn, priv, type, dev, fmt, ##args) 5274 #define netif_notice(priv, type, dev, fmt, args...) \ 5275 netif_level(notice, priv, type, dev, fmt, ##args) 5276 #define netif_info(priv, type, dev, fmt, args...) \ 5277 netif_level(info, priv, type, dev, fmt, ##args) 5278 5279 #if defined(CONFIG_DYNAMIC_DEBUG) || \ 5280 (defined(CONFIG_DYNAMIC_DEBUG_CORE) && defined(DYNAMIC_DEBUG_MODULE)) 5281 #define netif_dbg(priv, type, netdev, format, args...) \ 5282 do { \ 5283 if (netif_msg_##type(priv)) \ 5284 dynamic_netdev_dbg(netdev, format, ##args); \ 5285 } while (0) 5286 #elif defined(DEBUG) 5287 #define netif_dbg(priv, type, dev, format, args...) \ 5288 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args) 5289 #else 5290 #define netif_dbg(priv, type, dev, format, args...) \ 5291 ({ \ 5292 if (0) \ 5293 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \ 5294 0; \ 5295 }) 5296 #endif 5297 5298 /* if @cond then downgrade to debug, else print at @level */ 5299 #define netif_cond_dbg(priv, type, netdev, cond, level, fmt, args...) \ 5300 do { \ 5301 if (cond) \ 5302 netif_dbg(priv, type, netdev, fmt, ##args); \ 5303 else \ 5304 netif_ ## level(priv, type, netdev, fmt, ##args); \ 5305 } while (0) 5306 5307 #if defined(VERBOSE_DEBUG) 5308 #define netif_vdbg netif_dbg 5309 #else 5310 #define netif_vdbg(priv, type, dev, format, args...) \ 5311 ({ \ 5312 if (0) \ 5313 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \ 5314 0; \ 5315 }) 5316 #endif 5317 5318 /* 5319 * The list of packet types we will receive (as opposed to discard) 5320 * and the routines to invoke. 5321 * 5322 * Why 16. Because with 16 the only overlap we get on a hash of the 5323 * low nibble of the protocol value is RARP/SNAP/X.25. 5324 * 5325 * 0800 IP 5326 * 0001 802.3 5327 * 0002 AX.25 5328 * 0004 802.2 5329 * 8035 RARP 5330 * 0005 SNAP 5331 * 0805 X.25 5332 * 0806 ARP 5333 * 8137 IPX 5334 * 0009 Localtalk 5335 * 86DD IPv6 5336 */ 5337 #define PTYPE_HASH_SIZE (16) 5338 #define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1) 5339 5340 extern struct list_head ptype_all __read_mostly; 5341 extern struct list_head ptype_base[PTYPE_HASH_SIZE] __read_mostly; 5342 5343 extern struct net_device *blackhole_netdev; 5344 5345 #endif /* _LINUX_NETDEVICE_H */ 5346