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 NAPI_STATE_SCHED_THREADED, /* Napi is currently scheduled in threaded mode */ 364 }; 365 366 enum { 367 NAPIF_STATE_SCHED = BIT(NAPI_STATE_SCHED), 368 NAPIF_STATE_MISSED = BIT(NAPI_STATE_MISSED), 369 NAPIF_STATE_DISABLE = BIT(NAPI_STATE_DISABLE), 370 NAPIF_STATE_NPSVC = BIT(NAPI_STATE_NPSVC), 371 NAPIF_STATE_LISTED = BIT(NAPI_STATE_LISTED), 372 NAPIF_STATE_NO_BUSY_POLL = BIT(NAPI_STATE_NO_BUSY_POLL), 373 NAPIF_STATE_IN_BUSY_POLL = BIT(NAPI_STATE_IN_BUSY_POLL), 374 NAPIF_STATE_PREFER_BUSY_POLL = BIT(NAPI_STATE_PREFER_BUSY_POLL), 375 NAPIF_STATE_THREADED = BIT(NAPI_STATE_THREADED), 376 NAPIF_STATE_SCHED_THREADED = BIT(NAPI_STATE_SCHED_THREADED), 377 }; 378 379 enum gro_result { 380 GRO_MERGED, 381 GRO_MERGED_FREE, 382 GRO_HELD, 383 GRO_NORMAL, 384 GRO_CONSUMED, 385 }; 386 typedef enum gro_result gro_result_t; 387 388 /* 389 * enum rx_handler_result - Possible return values for rx_handlers. 390 * @RX_HANDLER_CONSUMED: skb was consumed by rx_handler, do not process it 391 * further. 392 * @RX_HANDLER_ANOTHER: Do another round in receive path. This is indicated in 393 * case skb->dev was changed by rx_handler. 394 * @RX_HANDLER_EXACT: Force exact delivery, no wildcard. 395 * @RX_HANDLER_PASS: Do nothing, pass the skb as if no rx_handler was called. 396 * 397 * rx_handlers are functions called from inside __netif_receive_skb(), to do 398 * special processing of the skb, prior to delivery to protocol handlers. 399 * 400 * Currently, a net_device can only have a single rx_handler registered. Trying 401 * to register a second rx_handler will return -EBUSY. 402 * 403 * To register a rx_handler on a net_device, use netdev_rx_handler_register(). 404 * To unregister a rx_handler on a net_device, use 405 * netdev_rx_handler_unregister(). 406 * 407 * Upon return, rx_handler is expected to tell __netif_receive_skb() what to 408 * do with the skb. 409 * 410 * If the rx_handler consumed the skb in some way, it should return 411 * RX_HANDLER_CONSUMED. This is appropriate when the rx_handler arranged for 412 * the skb to be delivered in some other way. 413 * 414 * If the rx_handler changed skb->dev, to divert the skb to another 415 * net_device, it should return RX_HANDLER_ANOTHER. The rx_handler for the 416 * new device will be called if it exists. 417 * 418 * If the rx_handler decides the skb should be ignored, it should return 419 * RX_HANDLER_EXACT. The skb will only be delivered to protocol handlers that 420 * are registered on exact device (ptype->dev == skb->dev). 421 * 422 * If the rx_handler didn't change skb->dev, but wants the skb to be normally 423 * delivered, it should return RX_HANDLER_PASS. 424 * 425 * A device without a registered rx_handler will behave as if rx_handler 426 * returned RX_HANDLER_PASS. 427 */ 428 429 enum rx_handler_result { 430 RX_HANDLER_CONSUMED, 431 RX_HANDLER_ANOTHER, 432 RX_HANDLER_EXACT, 433 RX_HANDLER_PASS, 434 }; 435 typedef enum rx_handler_result rx_handler_result_t; 436 typedef rx_handler_result_t rx_handler_func_t(struct sk_buff **pskb); 437 438 void __napi_schedule(struct napi_struct *n); 439 void __napi_schedule_irqoff(struct napi_struct *n); 440 441 static inline bool napi_disable_pending(struct napi_struct *n) 442 { 443 return test_bit(NAPI_STATE_DISABLE, &n->state); 444 } 445 446 static inline bool napi_prefer_busy_poll(struct napi_struct *n) 447 { 448 return test_bit(NAPI_STATE_PREFER_BUSY_POLL, &n->state); 449 } 450 451 bool napi_schedule_prep(struct napi_struct *n); 452 453 /** 454 * napi_schedule - schedule NAPI poll 455 * @n: NAPI context 456 * 457 * Schedule NAPI poll routine to be called if it is not already 458 * running. 459 */ 460 static inline void napi_schedule(struct napi_struct *n) 461 { 462 if (napi_schedule_prep(n)) 463 __napi_schedule(n); 464 } 465 466 /** 467 * napi_schedule_irqoff - schedule NAPI poll 468 * @n: NAPI context 469 * 470 * Variant of napi_schedule(), assuming hard irqs are masked. 471 */ 472 static inline void napi_schedule_irqoff(struct napi_struct *n) 473 { 474 if (napi_schedule_prep(n)) 475 __napi_schedule_irqoff(n); 476 } 477 478 /* Try to reschedule poll. Called by dev->poll() after napi_complete(). */ 479 static inline bool napi_reschedule(struct napi_struct *napi) 480 { 481 if (napi_schedule_prep(napi)) { 482 __napi_schedule(napi); 483 return true; 484 } 485 return false; 486 } 487 488 bool napi_complete_done(struct napi_struct *n, int work_done); 489 /** 490 * napi_complete - NAPI processing complete 491 * @n: NAPI context 492 * 493 * Mark NAPI processing as complete. 494 * Consider using napi_complete_done() instead. 495 * Return false if device should avoid rearming interrupts. 496 */ 497 static inline bool napi_complete(struct napi_struct *n) 498 { 499 return napi_complete_done(n, 0); 500 } 501 502 int dev_set_threaded(struct net_device *dev, bool threaded); 503 504 /** 505 * napi_disable - prevent NAPI from scheduling 506 * @n: NAPI context 507 * 508 * Stop NAPI from being scheduled on this context. 509 * Waits till any outstanding processing completes. 510 */ 511 void napi_disable(struct napi_struct *n); 512 513 void napi_enable(struct napi_struct *n); 514 515 /** 516 * napi_synchronize - wait until NAPI is not running 517 * @n: NAPI context 518 * 519 * Wait until NAPI is done being scheduled on this context. 520 * Waits till any outstanding processing completes but 521 * does not disable future activations. 522 */ 523 static inline void napi_synchronize(const struct napi_struct *n) 524 { 525 if (IS_ENABLED(CONFIG_SMP)) 526 while (test_bit(NAPI_STATE_SCHED, &n->state)) 527 msleep(1); 528 else 529 barrier(); 530 } 531 532 /** 533 * napi_if_scheduled_mark_missed - if napi is running, set the 534 * NAPIF_STATE_MISSED 535 * @n: NAPI context 536 * 537 * If napi is running, set the NAPIF_STATE_MISSED, and return true if 538 * NAPI is scheduled. 539 **/ 540 static inline bool napi_if_scheduled_mark_missed(struct napi_struct *n) 541 { 542 unsigned long val, new; 543 544 do { 545 val = READ_ONCE(n->state); 546 if (val & NAPIF_STATE_DISABLE) 547 return true; 548 549 if (!(val & NAPIF_STATE_SCHED)) 550 return false; 551 552 new = val | NAPIF_STATE_MISSED; 553 } while (cmpxchg(&n->state, val, new) != val); 554 555 return true; 556 } 557 558 enum netdev_queue_state_t { 559 __QUEUE_STATE_DRV_XOFF, 560 __QUEUE_STATE_STACK_XOFF, 561 __QUEUE_STATE_FROZEN, 562 }; 563 564 #define QUEUE_STATE_DRV_XOFF (1 << __QUEUE_STATE_DRV_XOFF) 565 #define QUEUE_STATE_STACK_XOFF (1 << __QUEUE_STATE_STACK_XOFF) 566 #define QUEUE_STATE_FROZEN (1 << __QUEUE_STATE_FROZEN) 567 568 #define QUEUE_STATE_ANY_XOFF (QUEUE_STATE_DRV_XOFF | QUEUE_STATE_STACK_XOFF) 569 #define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \ 570 QUEUE_STATE_FROZEN) 571 #define QUEUE_STATE_DRV_XOFF_OR_FROZEN (QUEUE_STATE_DRV_XOFF | \ 572 QUEUE_STATE_FROZEN) 573 574 /* 575 * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue. The 576 * netif_tx_* functions below are used to manipulate this flag. The 577 * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit 578 * queue independently. The netif_xmit_*stopped functions below are called 579 * to check if the queue has been stopped by the driver or stack (either 580 * of the XOFF bits are set in the state). Drivers should not need to call 581 * netif_xmit*stopped functions, they should only be using netif_tx_*. 582 */ 583 584 struct netdev_queue { 585 /* 586 * read-mostly part 587 */ 588 struct net_device *dev; 589 struct Qdisc __rcu *qdisc; 590 struct Qdisc *qdisc_sleeping; 591 #ifdef CONFIG_SYSFS 592 struct kobject kobj; 593 #endif 594 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA) 595 int numa_node; 596 #endif 597 unsigned long tx_maxrate; 598 /* 599 * Number of TX timeouts for this queue 600 * (/sys/class/net/DEV/Q/trans_timeout) 601 */ 602 unsigned long trans_timeout; 603 604 /* Subordinate device that the queue has been assigned to */ 605 struct net_device *sb_dev; 606 #ifdef CONFIG_XDP_SOCKETS 607 struct xsk_buff_pool *pool; 608 #endif 609 /* 610 * write-mostly part 611 */ 612 spinlock_t _xmit_lock ____cacheline_aligned_in_smp; 613 int xmit_lock_owner; 614 /* 615 * Time (in jiffies) of last Tx 616 */ 617 unsigned long trans_start; 618 619 unsigned long state; 620 621 #ifdef CONFIG_BQL 622 struct dql dql; 623 #endif 624 } ____cacheline_aligned_in_smp; 625 626 extern int sysctl_fb_tunnels_only_for_init_net; 627 extern int sysctl_devconf_inherit_init_net; 628 629 /* 630 * sysctl_fb_tunnels_only_for_init_net == 0 : For all netns 631 * == 1 : For initns only 632 * == 2 : For none. 633 */ 634 static inline bool net_has_fallback_tunnels(const struct net *net) 635 { 636 return !IS_ENABLED(CONFIG_SYSCTL) || 637 !sysctl_fb_tunnels_only_for_init_net || 638 (net == &init_net && sysctl_fb_tunnels_only_for_init_net == 1); 639 } 640 641 static inline int netdev_queue_numa_node_read(const struct netdev_queue *q) 642 { 643 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA) 644 return q->numa_node; 645 #else 646 return NUMA_NO_NODE; 647 #endif 648 } 649 650 static inline void netdev_queue_numa_node_write(struct netdev_queue *q, int node) 651 { 652 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA) 653 q->numa_node = node; 654 #endif 655 } 656 657 #ifdef CONFIG_RPS 658 /* 659 * This structure holds an RPS map which can be of variable length. The 660 * map is an array of CPUs. 661 */ 662 struct rps_map { 663 unsigned int len; 664 struct rcu_head rcu; 665 u16 cpus[]; 666 }; 667 #define RPS_MAP_SIZE(_num) (sizeof(struct rps_map) + ((_num) * sizeof(u16))) 668 669 /* 670 * The rps_dev_flow structure contains the mapping of a flow to a CPU, the 671 * tail pointer for that CPU's input queue at the time of last enqueue, and 672 * a hardware filter index. 673 */ 674 struct rps_dev_flow { 675 u16 cpu; 676 u16 filter; 677 unsigned int last_qtail; 678 }; 679 #define RPS_NO_FILTER 0xffff 680 681 /* 682 * The rps_dev_flow_table structure contains a table of flow mappings. 683 */ 684 struct rps_dev_flow_table { 685 unsigned int mask; 686 struct rcu_head rcu; 687 struct rps_dev_flow flows[]; 688 }; 689 #define RPS_DEV_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_dev_flow_table) + \ 690 ((_num) * sizeof(struct rps_dev_flow))) 691 692 /* 693 * The rps_sock_flow_table contains mappings of flows to the last CPU 694 * on which they were processed by the application (set in recvmsg). 695 * Each entry is a 32bit value. Upper part is the high-order bits 696 * of flow hash, lower part is CPU number. 697 * rps_cpu_mask is used to partition the space, depending on number of 698 * possible CPUs : rps_cpu_mask = roundup_pow_of_two(nr_cpu_ids) - 1 699 * For example, if 64 CPUs are possible, rps_cpu_mask = 0x3f, 700 * meaning we use 32-6=26 bits for the hash. 701 */ 702 struct rps_sock_flow_table { 703 u32 mask; 704 705 u32 ents[] ____cacheline_aligned_in_smp; 706 }; 707 #define RPS_SOCK_FLOW_TABLE_SIZE(_num) (offsetof(struct rps_sock_flow_table, ents[_num])) 708 709 #define RPS_NO_CPU 0xffff 710 711 extern u32 rps_cpu_mask; 712 extern struct rps_sock_flow_table __rcu *rps_sock_flow_table; 713 714 static inline void rps_record_sock_flow(struct rps_sock_flow_table *table, 715 u32 hash) 716 { 717 if (table && hash) { 718 unsigned int index = hash & table->mask; 719 u32 val = hash & ~rps_cpu_mask; 720 721 /* We only give a hint, preemption can change CPU under us */ 722 val |= raw_smp_processor_id(); 723 724 if (table->ents[index] != val) 725 table->ents[index] = val; 726 } 727 } 728 729 #ifdef CONFIG_RFS_ACCEL 730 bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index, u32 flow_id, 731 u16 filter_id); 732 #endif 733 #endif /* CONFIG_RPS */ 734 735 /* This structure contains an instance of an RX queue. */ 736 struct netdev_rx_queue { 737 #ifdef CONFIG_RPS 738 struct rps_map __rcu *rps_map; 739 struct rps_dev_flow_table __rcu *rps_flow_table; 740 #endif 741 struct kobject kobj; 742 struct net_device *dev; 743 struct xdp_rxq_info xdp_rxq; 744 #ifdef CONFIG_XDP_SOCKETS 745 struct xsk_buff_pool *pool; 746 #endif 747 } ____cacheline_aligned_in_smp; 748 749 /* 750 * RX queue sysfs structures and functions. 751 */ 752 struct rx_queue_attribute { 753 struct attribute attr; 754 ssize_t (*show)(struct netdev_rx_queue *queue, char *buf); 755 ssize_t (*store)(struct netdev_rx_queue *queue, 756 const char *buf, size_t len); 757 }; 758 759 /* XPS map type and offset of the xps map within net_device->xps_maps[]. */ 760 enum xps_map_type { 761 XPS_CPUS = 0, 762 XPS_RXQS, 763 XPS_MAPS_MAX, 764 }; 765 766 #ifdef CONFIG_XPS 767 /* 768 * This structure holds an XPS map which can be of variable length. The 769 * map is an array of queues. 770 */ 771 struct xps_map { 772 unsigned int len; 773 unsigned int alloc_len; 774 struct rcu_head rcu; 775 u16 queues[]; 776 }; 777 #define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16))) 778 #define XPS_MIN_MAP_ALLOC ((L1_CACHE_ALIGN(offsetof(struct xps_map, queues[1])) \ 779 - sizeof(struct xps_map)) / sizeof(u16)) 780 781 /* 782 * This structure holds all XPS maps for device. Maps are indexed by CPU. 783 * 784 * We keep track of the number of cpus/rxqs used when the struct is allocated, 785 * in nr_ids. This will help not accessing out-of-bound memory. 786 * 787 * We keep track of the number of traffic classes used when the struct is 788 * allocated, in num_tc. This will be used to navigate the maps, to ensure we're 789 * not crossing its upper bound, as the original dev->num_tc can be updated in 790 * the meantime. 791 */ 792 struct xps_dev_maps { 793 struct rcu_head rcu; 794 unsigned int nr_ids; 795 s16 num_tc; 796 struct xps_map __rcu *attr_map[]; /* Either CPUs map or RXQs map */ 797 }; 798 799 #define XPS_CPU_DEV_MAPS_SIZE(_tcs) (sizeof(struct xps_dev_maps) + \ 800 (nr_cpu_ids * (_tcs) * sizeof(struct xps_map *))) 801 802 #define XPS_RXQ_DEV_MAPS_SIZE(_tcs, _rxqs) (sizeof(struct xps_dev_maps) +\ 803 (_rxqs * (_tcs) * sizeof(struct xps_map *))) 804 805 #endif /* CONFIG_XPS */ 806 807 #define TC_MAX_QUEUE 16 808 #define TC_BITMASK 15 809 /* HW offloaded queuing disciplines txq count and offset maps */ 810 struct netdev_tc_txq { 811 u16 count; 812 u16 offset; 813 }; 814 815 #if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE) 816 /* 817 * This structure is to hold information about the device 818 * configured to run FCoE protocol stack. 819 */ 820 struct netdev_fcoe_hbainfo { 821 char manufacturer[64]; 822 char serial_number[64]; 823 char hardware_version[64]; 824 char driver_version[64]; 825 char optionrom_version[64]; 826 char firmware_version[64]; 827 char model[256]; 828 char model_description[256]; 829 }; 830 #endif 831 832 #define MAX_PHYS_ITEM_ID_LEN 32 833 834 /* This structure holds a unique identifier to identify some 835 * physical item (port for example) used by a netdevice. 836 */ 837 struct netdev_phys_item_id { 838 unsigned char id[MAX_PHYS_ITEM_ID_LEN]; 839 unsigned char id_len; 840 }; 841 842 static inline bool netdev_phys_item_id_same(struct netdev_phys_item_id *a, 843 struct netdev_phys_item_id *b) 844 { 845 return a->id_len == b->id_len && 846 memcmp(a->id, b->id, a->id_len) == 0; 847 } 848 849 typedef u16 (*select_queue_fallback_t)(struct net_device *dev, 850 struct sk_buff *skb, 851 struct net_device *sb_dev); 852 853 enum net_device_path_type { 854 DEV_PATH_ETHERNET = 0, 855 DEV_PATH_VLAN, 856 DEV_PATH_BRIDGE, 857 DEV_PATH_PPPOE, 858 DEV_PATH_DSA, 859 }; 860 861 struct net_device_path { 862 enum net_device_path_type type; 863 const struct net_device *dev; 864 union { 865 struct { 866 u16 id; 867 __be16 proto; 868 u8 h_dest[ETH_ALEN]; 869 } encap; 870 struct { 871 enum { 872 DEV_PATH_BR_VLAN_KEEP, 873 DEV_PATH_BR_VLAN_TAG, 874 DEV_PATH_BR_VLAN_UNTAG, 875 DEV_PATH_BR_VLAN_UNTAG_HW, 876 } vlan_mode; 877 u16 vlan_id; 878 __be16 vlan_proto; 879 } bridge; 880 struct { 881 int port; 882 u16 proto; 883 } dsa; 884 }; 885 }; 886 887 #define NET_DEVICE_PATH_STACK_MAX 5 888 #define NET_DEVICE_PATH_VLAN_MAX 2 889 890 struct net_device_path_stack { 891 int num_paths; 892 struct net_device_path path[NET_DEVICE_PATH_STACK_MAX]; 893 }; 894 895 struct net_device_path_ctx { 896 const struct net_device *dev; 897 const u8 *daddr; 898 899 int num_vlans; 900 struct { 901 u16 id; 902 __be16 proto; 903 } vlan[NET_DEVICE_PATH_VLAN_MAX]; 904 }; 905 906 enum tc_setup_type { 907 TC_SETUP_QDISC_MQPRIO, 908 TC_SETUP_CLSU32, 909 TC_SETUP_CLSFLOWER, 910 TC_SETUP_CLSMATCHALL, 911 TC_SETUP_CLSBPF, 912 TC_SETUP_BLOCK, 913 TC_SETUP_QDISC_CBS, 914 TC_SETUP_QDISC_RED, 915 TC_SETUP_QDISC_PRIO, 916 TC_SETUP_QDISC_MQ, 917 TC_SETUP_QDISC_ETF, 918 TC_SETUP_ROOT_QDISC, 919 TC_SETUP_QDISC_GRED, 920 TC_SETUP_QDISC_TAPRIO, 921 TC_SETUP_FT, 922 TC_SETUP_QDISC_ETS, 923 TC_SETUP_QDISC_TBF, 924 TC_SETUP_QDISC_FIFO, 925 TC_SETUP_QDISC_HTB, 926 }; 927 928 /* These structures hold the attributes of bpf state that are being passed 929 * to the netdevice through the bpf op. 930 */ 931 enum bpf_netdev_command { 932 /* Set or clear a bpf program used in the earliest stages of packet 933 * rx. The prog will have been loaded as BPF_PROG_TYPE_XDP. The callee 934 * is responsible for calling bpf_prog_put on any old progs that are 935 * stored. In case of error, the callee need not release the new prog 936 * reference, but on success it takes ownership and must bpf_prog_put 937 * when it is no longer used. 938 */ 939 XDP_SETUP_PROG, 940 XDP_SETUP_PROG_HW, 941 /* BPF program for offload callbacks, invoked at program load time. */ 942 BPF_OFFLOAD_MAP_ALLOC, 943 BPF_OFFLOAD_MAP_FREE, 944 XDP_SETUP_XSK_POOL, 945 }; 946 947 struct bpf_prog_offload_ops; 948 struct netlink_ext_ack; 949 struct xdp_umem; 950 struct xdp_dev_bulk_queue; 951 struct bpf_xdp_link; 952 953 enum bpf_xdp_mode { 954 XDP_MODE_SKB = 0, 955 XDP_MODE_DRV = 1, 956 XDP_MODE_HW = 2, 957 __MAX_XDP_MODE 958 }; 959 960 struct bpf_xdp_entity { 961 struct bpf_prog *prog; 962 struct bpf_xdp_link *link; 963 }; 964 965 struct netdev_bpf { 966 enum bpf_netdev_command command; 967 union { 968 /* XDP_SETUP_PROG */ 969 struct { 970 u32 flags; 971 struct bpf_prog *prog; 972 struct netlink_ext_ack *extack; 973 }; 974 /* BPF_OFFLOAD_MAP_ALLOC, BPF_OFFLOAD_MAP_FREE */ 975 struct { 976 struct bpf_offloaded_map *offmap; 977 }; 978 /* XDP_SETUP_XSK_POOL */ 979 struct { 980 struct xsk_buff_pool *pool; 981 u16 queue_id; 982 } xsk; 983 }; 984 }; 985 986 /* Flags for ndo_xsk_wakeup. */ 987 #define XDP_WAKEUP_RX (1 << 0) 988 #define XDP_WAKEUP_TX (1 << 1) 989 990 #ifdef CONFIG_XFRM_OFFLOAD 991 struct xfrmdev_ops { 992 int (*xdo_dev_state_add) (struct xfrm_state *x); 993 void (*xdo_dev_state_delete) (struct xfrm_state *x); 994 void (*xdo_dev_state_free) (struct xfrm_state *x); 995 bool (*xdo_dev_offload_ok) (struct sk_buff *skb, 996 struct xfrm_state *x); 997 void (*xdo_dev_state_advance_esn) (struct xfrm_state *x); 998 }; 999 #endif 1000 1001 struct dev_ifalias { 1002 struct rcu_head rcuhead; 1003 char ifalias[]; 1004 }; 1005 1006 struct devlink; 1007 struct tlsdev_ops; 1008 1009 struct netdev_name_node { 1010 struct hlist_node hlist; 1011 struct list_head list; 1012 struct net_device *dev; 1013 const char *name; 1014 }; 1015 1016 int netdev_name_node_alt_create(struct net_device *dev, const char *name); 1017 int netdev_name_node_alt_destroy(struct net_device *dev, const char *name); 1018 1019 struct netdev_net_notifier { 1020 struct list_head list; 1021 struct notifier_block *nb; 1022 }; 1023 1024 /* 1025 * This structure defines the management hooks for network devices. 1026 * The following hooks can be defined; unless noted otherwise, they are 1027 * optional and can be filled with a null pointer. 1028 * 1029 * int (*ndo_init)(struct net_device *dev); 1030 * This function is called once when a network device is registered. 1031 * The network device can use this for any late stage initialization 1032 * or semantic validation. It can fail with an error code which will 1033 * be propagated back to register_netdev. 1034 * 1035 * void (*ndo_uninit)(struct net_device *dev); 1036 * This function is called when device is unregistered or when registration 1037 * fails. It is not called if init fails. 1038 * 1039 * int (*ndo_open)(struct net_device *dev); 1040 * This function is called when a network device transitions to the up 1041 * state. 1042 * 1043 * int (*ndo_stop)(struct net_device *dev); 1044 * This function is called when a network device transitions to the down 1045 * state. 1046 * 1047 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb, 1048 * struct net_device *dev); 1049 * Called when a packet needs to be transmitted. 1050 * Returns NETDEV_TX_OK. Can return NETDEV_TX_BUSY, but you should stop 1051 * the queue before that can happen; it's for obsolete devices and weird 1052 * corner cases, but the stack really does a non-trivial amount 1053 * of useless work if you return NETDEV_TX_BUSY. 1054 * Required; cannot be NULL. 1055 * 1056 * netdev_features_t (*ndo_features_check)(struct sk_buff *skb, 1057 * struct net_device *dev 1058 * netdev_features_t features); 1059 * Called by core transmit path to determine if device is capable of 1060 * performing offload operations on a given packet. This is to give 1061 * the device an opportunity to implement any restrictions that cannot 1062 * be otherwise expressed by feature flags. The check is called with 1063 * the set of features that the stack has calculated and it returns 1064 * those the driver believes to be appropriate. 1065 * 1066 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb, 1067 * struct net_device *sb_dev); 1068 * Called to decide which queue to use when device supports multiple 1069 * transmit queues. 1070 * 1071 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags); 1072 * This function is called to allow device receiver to make 1073 * changes to configuration when multicast or promiscuous is enabled. 1074 * 1075 * void (*ndo_set_rx_mode)(struct net_device *dev); 1076 * This function is called device changes address list filtering. 1077 * If driver handles unicast address filtering, it should set 1078 * IFF_UNICAST_FLT in its priv_flags. 1079 * 1080 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr); 1081 * This function is called when the Media Access Control address 1082 * needs to be changed. If this interface is not defined, the 1083 * MAC address can not be changed. 1084 * 1085 * int (*ndo_validate_addr)(struct net_device *dev); 1086 * Test if Media Access Control address is valid for the device. 1087 * 1088 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd); 1089 * Called when a user requests an ioctl which can't be handled by 1090 * the generic interface code. If not defined ioctls return 1091 * not supported error code. 1092 * 1093 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map); 1094 * Used to set network devices bus interface parameters. This interface 1095 * is retained for legacy reasons; new devices should use the bus 1096 * interface (PCI) for low level management. 1097 * 1098 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu); 1099 * Called when a user wants to change the Maximum Transfer Unit 1100 * of a device. 1101 * 1102 * void (*ndo_tx_timeout)(struct net_device *dev, unsigned int txqueue); 1103 * Callback used when the transmitter has not made any progress 1104 * for dev->watchdog ticks. 1105 * 1106 * void (*ndo_get_stats64)(struct net_device *dev, 1107 * struct rtnl_link_stats64 *storage); 1108 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev); 1109 * Called when a user wants to get the network device usage 1110 * statistics. Drivers must do one of the following: 1111 * 1. Define @ndo_get_stats64 to fill in a zero-initialised 1112 * rtnl_link_stats64 structure passed by the caller. 1113 * 2. Define @ndo_get_stats to update a net_device_stats structure 1114 * (which should normally be dev->stats) and return a pointer to 1115 * it. The structure may be changed asynchronously only if each 1116 * field is written atomically. 1117 * 3. Update dev->stats asynchronously and atomically, and define 1118 * neither operation. 1119 * 1120 * bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id) 1121 * Return true if this device supports offload stats of this attr_id. 1122 * 1123 * int (*ndo_get_offload_stats)(int attr_id, const struct net_device *dev, 1124 * void *attr_data) 1125 * Get statistics for offload operations by attr_id. Write it into the 1126 * attr_data pointer. 1127 * 1128 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16 vid); 1129 * If device supports VLAN filtering this function is called when a 1130 * VLAN id is registered. 1131 * 1132 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, __be16 proto, u16 vid); 1133 * If device supports VLAN filtering this function is called when a 1134 * VLAN id is unregistered. 1135 * 1136 * void (*ndo_poll_controller)(struct net_device *dev); 1137 * 1138 * SR-IOV management functions. 1139 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac); 1140 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan, 1141 * u8 qos, __be16 proto); 1142 * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate, 1143 * int max_tx_rate); 1144 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting); 1145 * int (*ndo_set_vf_trust)(struct net_device *dev, int vf, bool setting); 1146 * int (*ndo_get_vf_config)(struct net_device *dev, 1147 * int vf, struct ifla_vf_info *ivf); 1148 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state); 1149 * int (*ndo_set_vf_port)(struct net_device *dev, int vf, 1150 * struct nlattr *port[]); 1151 * 1152 * Enable or disable the VF ability to query its RSS Redirection Table and 1153 * Hash Key. This is needed since on some devices VF share this information 1154 * with PF and querying it may introduce a theoretical security risk. 1155 * int (*ndo_set_vf_rss_query_en)(struct net_device *dev, int vf, bool setting); 1156 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb); 1157 * int (*ndo_setup_tc)(struct net_device *dev, enum tc_setup_type type, 1158 * void *type_data); 1159 * Called to setup any 'tc' scheduler, classifier or action on @dev. 1160 * This is always called from the stack with the rtnl lock held and netif 1161 * tx queues stopped. This allows the netdevice to perform queue 1162 * management safely. 1163 * 1164 * Fiber Channel over Ethernet (FCoE) offload functions. 1165 * int (*ndo_fcoe_enable)(struct net_device *dev); 1166 * Called when the FCoE protocol stack wants to start using LLD for FCoE 1167 * so the underlying device can perform whatever needed configuration or 1168 * initialization to support acceleration of FCoE traffic. 1169 * 1170 * int (*ndo_fcoe_disable)(struct net_device *dev); 1171 * Called when the FCoE protocol stack wants to stop using LLD for FCoE 1172 * so the underlying device can perform whatever needed clean-ups to 1173 * stop supporting acceleration of FCoE traffic. 1174 * 1175 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid, 1176 * struct scatterlist *sgl, unsigned int sgc); 1177 * Called when the FCoE Initiator wants to initialize an I/O that 1178 * is a possible candidate for Direct Data Placement (DDP). The LLD can 1179 * perform necessary setup and returns 1 to indicate the device is set up 1180 * successfully to perform DDP on this I/O, otherwise this returns 0. 1181 * 1182 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid); 1183 * Called when the FCoE Initiator/Target is done with the DDPed I/O as 1184 * indicated by the FC exchange id 'xid', so the underlying device can 1185 * clean up and reuse resources for later DDP requests. 1186 * 1187 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid, 1188 * struct scatterlist *sgl, unsigned int sgc); 1189 * Called when the FCoE Target wants to initialize an I/O that 1190 * is a possible candidate for Direct Data Placement (DDP). The LLD can 1191 * perform necessary setup and returns 1 to indicate the device is set up 1192 * successfully to perform DDP on this I/O, otherwise this returns 0. 1193 * 1194 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev, 1195 * struct netdev_fcoe_hbainfo *hbainfo); 1196 * Called when the FCoE Protocol stack wants information on the underlying 1197 * device. This information is utilized by the FCoE protocol stack to 1198 * register attributes with Fiber Channel management service as per the 1199 * FC-GS Fabric Device Management Information(FDMI) specification. 1200 * 1201 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type); 1202 * Called when the underlying device wants to override default World Wide 1203 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own 1204 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE 1205 * protocol stack to use. 1206 * 1207 * RFS acceleration. 1208 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb, 1209 * u16 rxq_index, u32 flow_id); 1210 * Set hardware filter for RFS. rxq_index is the target queue index; 1211 * flow_id is a flow ID to be passed to rps_may_expire_flow() later. 1212 * Return the filter ID on success, or a negative error code. 1213 * 1214 * Slave management functions (for bridge, bonding, etc). 1215 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev); 1216 * Called to make another netdev an underling. 1217 * 1218 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev); 1219 * Called to release previously enslaved netdev. 1220 * 1221 * struct net_device *(*ndo_get_xmit_slave)(struct net_device *dev, 1222 * struct sk_buff *skb, 1223 * bool all_slaves); 1224 * Get the xmit slave of master device. If all_slaves is true, function 1225 * assume all the slaves can transmit. 1226 * 1227 * Feature/offload setting functions. 1228 * netdev_features_t (*ndo_fix_features)(struct net_device *dev, 1229 * netdev_features_t features); 1230 * Adjusts the requested feature flags according to device-specific 1231 * constraints, and returns the resulting flags. Must not modify 1232 * the device state. 1233 * 1234 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features); 1235 * Called to update device configuration to new features. Passed 1236 * feature set might be less than what was returned by ndo_fix_features()). 1237 * Must return >0 or -errno if it changed dev->features itself. 1238 * 1239 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[], 1240 * struct net_device *dev, 1241 * const unsigned char *addr, u16 vid, u16 flags, 1242 * struct netlink_ext_ack *extack); 1243 * Adds an FDB entry to dev for addr. 1244 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[], 1245 * struct net_device *dev, 1246 * const unsigned char *addr, u16 vid) 1247 * Deletes the FDB entry from dev coresponding to addr. 1248 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb, 1249 * struct net_device *dev, struct net_device *filter_dev, 1250 * int *idx) 1251 * Used to add FDB entries to dump requests. Implementers should add 1252 * entries to skb and update idx with the number of entries. 1253 * 1254 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh, 1255 * u16 flags, struct netlink_ext_ack *extack) 1256 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq, 1257 * struct net_device *dev, u32 filter_mask, 1258 * int nlflags) 1259 * int (*ndo_bridge_dellink)(struct net_device *dev, struct nlmsghdr *nlh, 1260 * u16 flags); 1261 * 1262 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier); 1263 * Called to change device carrier. Soft-devices (like dummy, team, etc) 1264 * which do not represent real hardware may define this to allow their 1265 * userspace components to manage their virtual carrier state. Devices 1266 * that determine carrier state from physical hardware properties (eg 1267 * network cables) or protocol-dependent mechanisms (eg 1268 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function. 1269 * 1270 * int (*ndo_get_phys_port_id)(struct net_device *dev, 1271 * struct netdev_phys_item_id *ppid); 1272 * Called to get ID of physical port of this device. If driver does 1273 * not implement this, it is assumed that the hw is not able to have 1274 * multiple net devices on single physical port. 1275 * 1276 * int (*ndo_get_port_parent_id)(struct net_device *dev, 1277 * struct netdev_phys_item_id *ppid) 1278 * Called to get the parent ID of the physical port of this device. 1279 * 1280 * void* (*ndo_dfwd_add_station)(struct net_device *pdev, 1281 * struct net_device *dev) 1282 * Called by upper layer devices to accelerate switching or other 1283 * station functionality into hardware. 'pdev is the lowerdev 1284 * to use for the offload and 'dev' is the net device that will 1285 * back the offload. Returns a pointer to the private structure 1286 * the upper layer will maintain. 1287 * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv) 1288 * Called by upper layer device to delete the station created 1289 * by 'ndo_dfwd_add_station'. 'pdev' is the net device backing 1290 * the station and priv is the structure returned by the add 1291 * operation. 1292 * int (*ndo_set_tx_maxrate)(struct net_device *dev, 1293 * int queue_index, u32 maxrate); 1294 * Called when a user wants to set a max-rate limitation of specific 1295 * TX queue. 1296 * int (*ndo_get_iflink)(const struct net_device *dev); 1297 * Called to get the iflink value of this device. 1298 * void (*ndo_change_proto_down)(struct net_device *dev, 1299 * bool proto_down); 1300 * This function is used to pass protocol port error state information 1301 * to the switch driver. The switch driver can react to the proto_down 1302 * by doing a phys down on the associated switch port. 1303 * int (*ndo_fill_metadata_dst)(struct net_device *dev, struct sk_buff *skb); 1304 * This function is used to get egress tunnel information for given skb. 1305 * This is useful for retrieving outer tunnel header parameters while 1306 * sampling packet. 1307 * void (*ndo_set_rx_headroom)(struct net_device *dev, int needed_headroom); 1308 * This function is used to specify the headroom that the skb must 1309 * consider when allocation skb during packet reception. Setting 1310 * appropriate rx headroom value allows avoiding skb head copy on 1311 * forward. Setting a negative value resets the rx headroom to the 1312 * default value. 1313 * int (*ndo_bpf)(struct net_device *dev, struct netdev_bpf *bpf); 1314 * This function is used to set or query state related to XDP on the 1315 * netdevice and manage BPF offload. See definition of 1316 * enum bpf_netdev_command for details. 1317 * int (*ndo_xdp_xmit)(struct net_device *dev, int n, struct xdp_frame **xdp, 1318 * u32 flags); 1319 * This function is used to submit @n XDP packets for transmit on a 1320 * netdevice. Returns number of frames successfully transmitted, frames 1321 * that got dropped are freed/returned via xdp_return_frame(). 1322 * Returns negative number, means general error invoking ndo, meaning 1323 * no frames were xmit'ed and core-caller will free all frames. 1324 * int (*ndo_xsk_wakeup)(struct net_device *dev, u32 queue_id, u32 flags); 1325 * This function is used to wake up the softirq, ksoftirqd or kthread 1326 * responsible for sending and/or receiving packets on a specific 1327 * queue id bound to an AF_XDP socket. The flags field specifies if 1328 * only RX, only Tx, or both should be woken up using the flags 1329 * XDP_WAKEUP_RX and XDP_WAKEUP_TX. 1330 * struct devlink_port *(*ndo_get_devlink_port)(struct net_device *dev); 1331 * Get devlink port instance associated with a given netdev. 1332 * Called with a reference on the netdevice and devlink locks only, 1333 * rtnl_lock is not held. 1334 * int (*ndo_tunnel_ctl)(struct net_device *dev, struct ip_tunnel_parm *p, 1335 * int cmd); 1336 * Add, change, delete or get information on an IPv4 tunnel. 1337 * struct net_device *(*ndo_get_peer_dev)(struct net_device *dev); 1338 * If a device is paired with a peer device, return the peer instance. 1339 * The caller must be under RCU read context. 1340 * int (*ndo_fill_forward_path)(struct net_device_path_ctx *ctx, struct net_device_path *path); 1341 * Get the forwarding path to reach the real device from the HW destination address 1342 */ 1343 struct net_device_ops { 1344 int (*ndo_init)(struct net_device *dev); 1345 void (*ndo_uninit)(struct net_device *dev); 1346 int (*ndo_open)(struct net_device *dev); 1347 int (*ndo_stop)(struct net_device *dev); 1348 netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb, 1349 struct net_device *dev); 1350 netdev_features_t (*ndo_features_check)(struct sk_buff *skb, 1351 struct net_device *dev, 1352 netdev_features_t features); 1353 u16 (*ndo_select_queue)(struct net_device *dev, 1354 struct sk_buff *skb, 1355 struct net_device *sb_dev); 1356 void (*ndo_change_rx_flags)(struct net_device *dev, 1357 int flags); 1358 void (*ndo_set_rx_mode)(struct net_device *dev); 1359 int (*ndo_set_mac_address)(struct net_device *dev, 1360 void *addr); 1361 int (*ndo_validate_addr)(struct net_device *dev); 1362 int (*ndo_do_ioctl)(struct net_device *dev, 1363 struct ifreq *ifr, int cmd); 1364 int (*ndo_set_config)(struct net_device *dev, 1365 struct ifmap *map); 1366 int (*ndo_change_mtu)(struct net_device *dev, 1367 int new_mtu); 1368 int (*ndo_neigh_setup)(struct net_device *dev, 1369 struct neigh_parms *); 1370 void (*ndo_tx_timeout) (struct net_device *dev, 1371 unsigned int txqueue); 1372 1373 void (*ndo_get_stats64)(struct net_device *dev, 1374 struct rtnl_link_stats64 *storage); 1375 bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id); 1376 int (*ndo_get_offload_stats)(int attr_id, 1377 const struct net_device *dev, 1378 void *attr_data); 1379 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev); 1380 1381 int (*ndo_vlan_rx_add_vid)(struct net_device *dev, 1382 __be16 proto, u16 vid); 1383 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, 1384 __be16 proto, u16 vid); 1385 #ifdef CONFIG_NET_POLL_CONTROLLER 1386 void (*ndo_poll_controller)(struct net_device *dev); 1387 int (*ndo_netpoll_setup)(struct net_device *dev, 1388 struct netpoll_info *info); 1389 void (*ndo_netpoll_cleanup)(struct net_device *dev); 1390 #endif 1391 int (*ndo_set_vf_mac)(struct net_device *dev, 1392 int queue, u8 *mac); 1393 int (*ndo_set_vf_vlan)(struct net_device *dev, 1394 int queue, u16 vlan, 1395 u8 qos, __be16 proto); 1396 int (*ndo_set_vf_rate)(struct net_device *dev, 1397 int vf, int min_tx_rate, 1398 int max_tx_rate); 1399 int (*ndo_set_vf_spoofchk)(struct net_device *dev, 1400 int vf, bool setting); 1401 int (*ndo_set_vf_trust)(struct net_device *dev, 1402 int vf, bool setting); 1403 int (*ndo_get_vf_config)(struct net_device *dev, 1404 int vf, 1405 struct ifla_vf_info *ivf); 1406 int (*ndo_set_vf_link_state)(struct net_device *dev, 1407 int vf, int link_state); 1408 int (*ndo_get_vf_stats)(struct net_device *dev, 1409 int vf, 1410 struct ifla_vf_stats 1411 *vf_stats); 1412 int (*ndo_set_vf_port)(struct net_device *dev, 1413 int vf, 1414 struct nlattr *port[]); 1415 int (*ndo_get_vf_port)(struct net_device *dev, 1416 int vf, struct sk_buff *skb); 1417 int (*ndo_get_vf_guid)(struct net_device *dev, 1418 int vf, 1419 struct ifla_vf_guid *node_guid, 1420 struct ifla_vf_guid *port_guid); 1421 int (*ndo_set_vf_guid)(struct net_device *dev, 1422 int vf, u64 guid, 1423 int guid_type); 1424 int (*ndo_set_vf_rss_query_en)( 1425 struct net_device *dev, 1426 int vf, bool setting); 1427 int (*ndo_setup_tc)(struct net_device *dev, 1428 enum tc_setup_type type, 1429 void *type_data); 1430 #if IS_ENABLED(CONFIG_FCOE) 1431 int (*ndo_fcoe_enable)(struct net_device *dev); 1432 int (*ndo_fcoe_disable)(struct net_device *dev); 1433 int (*ndo_fcoe_ddp_setup)(struct net_device *dev, 1434 u16 xid, 1435 struct scatterlist *sgl, 1436 unsigned int sgc); 1437 int (*ndo_fcoe_ddp_done)(struct net_device *dev, 1438 u16 xid); 1439 int (*ndo_fcoe_ddp_target)(struct net_device *dev, 1440 u16 xid, 1441 struct scatterlist *sgl, 1442 unsigned int sgc); 1443 int (*ndo_fcoe_get_hbainfo)(struct net_device *dev, 1444 struct netdev_fcoe_hbainfo *hbainfo); 1445 #endif 1446 1447 #if IS_ENABLED(CONFIG_LIBFCOE) 1448 #define NETDEV_FCOE_WWNN 0 1449 #define NETDEV_FCOE_WWPN 1 1450 int (*ndo_fcoe_get_wwn)(struct net_device *dev, 1451 u64 *wwn, int type); 1452 #endif 1453 1454 #ifdef CONFIG_RFS_ACCEL 1455 int (*ndo_rx_flow_steer)(struct net_device *dev, 1456 const struct sk_buff *skb, 1457 u16 rxq_index, 1458 u32 flow_id); 1459 #endif 1460 int (*ndo_add_slave)(struct net_device *dev, 1461 struct net_device *slave_dev, 1462 struct netlink_ext_ack *extack); 1463 int (*ndo_del_slave)(struct net_device *dev, 1464 struct net_device *slave_dev); 1465 struct net_device* (*ndo_get_xmit_slave)(struct net_device *dev, 1466 struct sk_buff *skb, 1467 bool all_slaves); 1468 struct net_device* (*ndo_sk_get_lower_dev)(struct net_device *dev, 1469 struct sock *sk); 1470 netdev_features_t (*ndo_fix_features)(struct net_device *dev, 1471 netdev_features_t features); 1472 int (*ndo_set_features)(struct net_device *dev, 1473 netdev_features_t features); 1474 int (*ndo_neigh_construct)(struct net_device *dev, 1475 struct neighbour *n); 1476 void (*ndo_neigh_destroy)(struct net_device *dev, 1477 struct neighbour *n); 1478 1479 int (*ndo_fdb_add)(struct ndmsg *ndm, 1480 struct nlattr *tb[], 1481 struct net_device *dev, 1482 const unsigned char *addr, 1483 u16 vid, 1484 u16 flags, 1485 struct netlink_ext_ack *extack); 1486 int (*ndo_fdb_del)(struct ndmsg *ndm, 1487 struct nlattr *tb[], 1488 struct net_device *dev, 1489 const unsigned char *addr, 1490 u16 vid); 1491 int (*ndo_fdb_dump)(struct sk_buff *skb, 1492 struct netlink_callback *cb, 1493 struct net_device *dev, 1494 struct net_device *filter_dev, 1495 int *idx); 1496 int (*ndo_fdb_get)(struct sk_buff *skb, 1497 struct nlattr *tb[], 1498 struct net_device *dev, 1499 const unsigned char *addr, 1500 u16 vid, u32 portid, u32 seq, 1501 struct netlink_ext_ack *extack); 1502 int (*ndo_bridge_setlink)(struct net_device *dev, 1503 struct nlmsghdr *nlh, 1504 u16 flags, 1505 struct netlink_ext_ack *extack); 1506 int (*ndo_bridge_getlink)(struct sk_buff *skb, 1507 u32 pid, u32 seq, 1508 struct net_device *dev, 1509 u32 filter_mask, 1510 int nlflags); 1511 int (*ndo_bridge_dellink)(struct net_device *dev, 1512 struct nlmsghdr *nlh, 1513 u16 flags); 1514 int (*ndo_change_carrier)(struct net_device *dev, 1515 bool new_carrier); 1516 int (*ndo_get_phys_port_id)(struct net_device *dev, 1517 struct netdev_phys_item_id *ppid); 1518 int (*ndo_get_port_parent_id)(struct net_device *dev, 1519 struct netdev_phys_item_id *ppid); 1520 int (*ndo_get_phys_port_name)(struct net_device *dev, 1521 char *name, size_t len); 1522 void* (*ndo_dfwd_add_station)(struct net_device *pdev, 1523 struct net_device *dev); 1524 void (*ndo_dfwd_del_station)(struct net_device *pdev, 1525 void *priv); 1526 1527 int (*ndo_set_tx_maxrate)(struct net_device *dev, 1528 int queue_index, 1529 u32 maxrate); 1530 int (*ndo_get_iflink)(const struct net_device *dev); 1531 int (*ndo_change_proto_down)(struct net_device *dev, 1532 bool proto_down); 1533 int (*ndo_fill_metadata_dst)(struct net_device *dev, 1534 struct sk_buff *skb); 1535 void (*ndo_set_rx_headroom)(struct net_device *dev, 1536 int needed_headroom); 1537 int (*ndo_bpf)(struct net_device *dev, 1538 struct netdev_bpf *bpf); 1539 int (*ndo_xdp_xmit)(struct net_device *dev, int n, 1540 struct xdp_frame **xdp, 1541 u32 flags); 1542 int (*ndo_xsk_wakeup)(struct net_device *dev, 1543 u32 queue_id, u32 flags); 1544 struct devlink_port * (*ndo_get_devlink_port)(struct net_device *dev); 1545 int (*ndo_tunnel_ctl)(struct net_device *dev, 1546 struct ip_tunnel_parm *p, int cmd); 1547 struct net_device * (*ndo_get_peer_dev)(struct net_device *dev); 1548 int (*ndo_fill_forward_path)(struct net_device_path_ctx *ctx, 1549 struct net_device_path *path); 1550 }; 1551 1552 /** 1553 * enum netdev_priv_flags - &struct net_device priv_flags 1554 * 1555 * These are the &struct net_device, they are only set internally 1556 * by drivers and used in the kernel. These flags are invisible to 1557 * userspace; this means that the order of these flags can change 1558 * during any kernel release. 1559 * 1560 * You should have a pretty good reason to be extending these flags. 1561 * 1562 * @IFF_802_1Q_VLAN: 802.1Q VLAN device 1563 * @IFF_EBRIDGE: Ethernet bridging device 1564 * @IFF_BONDING: bonding master or slave 1565 * @IFF_ISATAP: ISATAP interface (RFC4214) 1566 * @IFF_WAN_HDLC: WAN HDLC device 1567 * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to 1568 * release skb->dst 1569 * @IFF_DONT_BRIDGE: disallow bridging this ether dev 1570 * @IFF_DISABLE_NETPOLL: disable netpoll at run-time 1571 * @IFF_MACVLAN_PORT: device used as macvlan port 1572 * @IFF_BRIDGE_PORT: device used as bridge port 1573 * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port 1574 * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit 1575 * @IFF_UNICAST_FLT: Supports unicast filtering 1576 * @IFF_TEAM_PORT: device used as team port 1577 * @IFF_SUPP_NOFCS: device supports sending custom FCS 1578 * @IFF_LIVE_ADDR_CHANGE: device supports hardware address 1579 * change when it's running 1580 * @IFF_MACVLAN: Macvlan device 1581 * @IFF_XMIT_DST_RELEASE_PERM: IFF_XMIT_DST_RELEASE not taking into account 1582 * underlying stacked devices 1583 * @IFF_L3MDEV_MASTER: device is an L3 master device 1584 * @IFF_NO_QUEUE: device can run without qdisc attached 1585 * @IFF_OPENVSWITCH: device is a Open vSwitch master 1586 * @IFF_L3MDEV_SLAVE: device is enslaved to an L3 master device 1587 * @IFF_TEAM: device is a team device 1588 * @IFF_RXFH_CONFIGURED: device has had Rx Flow indirection table configured 1589 * @IFF_PHONY_HEADROOM: the headroom value is controlled by an external 1590 * entity (i.e. the master device for bridged veth) 1591 * @IFF_MACSEC: device is a MACsec device 1592 * @IFF_NO_RX_HANDLER: device doesn't support the rx_handler hook 1593 * @IFF_FAILOVER: device is a failover master device 1594 * @IFF_FAILOVER_SLAVE: device is lower dev of a failover master device 1595 * @IFF_L3MDEV_RX_HANDLER: only invoke the rx handler of L3 master device 1596 * @IFF_LIVE_RENAME_OK: rename is allowed while device is up and running 1597 * @IFF_TX_SKB_NO_LINEAR: device/driver is capable of xmitting frames with 1598 * skb_headlen(skb) == 0 (data starts from frag0) 1599 */ 1600 enum netdev_priv_flags { 1601 IFF_802_1Q_VLAN = 1<<0, 1602 IFF_EBRIDGE = 1<<1, 1603 IFF_BONDING = 1<<2, 1604 IFF_ISATAP = 1<<3, 1605 IFF_WAN_HDLC = 1<<4, 1606 IFF_XMIT_DST_RELEASE = 1<<5, 1607 IFF_DONT_BRIDGE = 1<<6, 1608 IFF_DISABLE_NETPOLL = 1<<7, 1609 IFF_MACVLAN_PORT = 1<<8, 1610 IFF_BRIDGE_PORT = 1<<9, 1611 IFF_OVS_DATAPATH = 1<<10, 1612 IFF_TX_SKB_SHARING = 1<<11, 1613 IFF_UNICAST_FLT = 1<<12, 1614 IFF_TEAM_PORT = 1<<13, 1615 IFF_SUPP_NOFCS = 1<<14, 1616 IFF_LIVE_ADDR_CHANGE = 1<<15, 1617 IFF_MACVLAN = 1<<16, 1618 IFF_XMIT_DST_RELEASE_PERM = 1<<17, 1619 IFF_L3MDEV_MASTER = 1<<18, 1620 IFF_NO_QUEUE = 1<<19, 1621 IFF_OPENVSWITCH = 1<<20, 1622 IFF_L3MDEV_SLAVE = 1<<21, 1623 IFF_TEAM = 1<<22, 1624 IFF_RXFH_CONFIGURED = 1<<23, 1625 IFF_PHONY_HEADROOM = 1<<24, 1626 IFF_MACSEC = 1<<25, 1627 IFF_NO_RX_HANDLER = 1<<26, 1628 IFF_FAILOVER = 1<<27, 1629 IFF_FAILOVER_SLAVE = 1<<28, 1630 IFF_L3MDEV_RX_HANDLER = 1<<29, 1631 IFF_LIVE_RENAME_OK = 1<<30, 1632 IFF_TX_SKB_NO_LINEAR = 1<<31, 1633 }; 1634 1635 #define IFF_802_1Q_VLAN IFF_802_1Q_VLAN 1636 #define IFF_EBRIDGE IFF_EBRIDGE 1637 #define IFF_BONDING IFF_BONDING 1638 #define IFF_ISATAP IFF_ISATAP 1639 #define IFF_WAN_HDLC IFF_WAN_HDLC 1640 #define IFF_XMIT_DST_RELEASE IFF_XMIT_DST_RELEASE 1641 #define IFF_DONT_BRIDGE IFF_DONT_BRIDGE 1642 #define IFF_DISABLE_NETPOLL IFF_DISABLE_NETPOLL 1643 #define IFF_MACVLAN_PORT IFF_MACVLAN_PORT 1644 #define IFF_BRIDGE_PORT IFF_BRIDGE_PORT 1645 #define IFF_OVS_DATAPATH IFF_OVS_DATAPATH 1646 #define IFF_TX_SKB_SHARING IFF_TX_SKB_SHARING 1647 #define IFF_UNICAST_FLT IFF_UNICAST_FLT 1648 #define IFF_TEAM_PORT IFF_TEAM_PORT 1649 #define IFF_SUPP_NOFCS IFF_SUPP_NOFCS 1650 #define IFF_LIVE_ADDR_CHANGE IFF_LIVE_ADDR_CHANGE 1651 #define IFF_MACVLAN IFF_MACVLAN 1652 #define IFF_XMIT_DST_RELEASE_PERM IFF_XMIT_DST_RELEASE_PERM 1653 #define IFF_L3MDEV_MASTER IFF_L3MDEV_MASTER 1654 #define IFF_NO_QUEUE IFF_NO_QUEUE 1655 #define IFF_OPENVSWITCH IFF_OPENVSWITCH 1656 #define IFF_L3MDEV_SLAVE IFF_L3MDEV_SLAVE 1657 #define IFF_TEAM IFF_TEAM 1658 #define IFF_RXFH_CONFIGURED IFF_RXFH_CONFIGURED 1659 #define IFF_PHONY_HEADROOM IFF_PHONY_HEADROOM 1660 #define IFF_MACSEC IFF_MACSEC 1661 #define IFF_NO_RX_HANDLER IFF_NO_RX_HANDLER 1662 #define IFF_FAILOVER IFF_FAILOVER 1663 #define IFF_FAILOVER_SLAVE IFF_FAILOVER_SLAVE 1664 #define IFF_L3MDEV_RX_HANDLER IFF_L3MDEV_RX_HANDLER 1665 #define IFF_LIVE_RENAME_OK IFF_LIVE_RENAME_OK 1666 #define IFF_TX_SKB_NO_LINEAR IFF_TX_SKB_NO_LINEAR 1667 1668 /* Specifies the type of the struct net_device::ml_priv pointer */ 1669 enum netdev_ml_priv_type { 1670 ML_PRIV_NONE, 1671 ML_PRIV_CAN, 1672 }; 1673 1674 /** 1675 * struct net_device - The DEVICE structure. 1676 * 1677 * Actually, this whole structure is a big mistake. It mixes I/O 1678 * data with strictly "high-level" data, and it has to know about 1679 * almost every data structure used in the INET module. 1680 * 1681 * @name: This is the first field of the "visible" part of this structure 1682 * (i.e. as seen by users in the "Space.c" file). It is the name 1683 * of the interface. 1684 * 1685 * @name_node: Name hashlist node 1686 * @ifalias: SNMP alias 1687 * @mem_end: Shared memory end 1688 * @mem_start: Shared memory start 1689 * @base_addr: Device I/O address 1690 * @irq: Device IRQ number 1691 * 1692 * @state: Generic network queuing layer state, see netdev_state_t 1693 * @dev_list: The global list of network devices 1694 * @napi_list: List entry used for polling NAPI devices 1695 * @unreg_list: List entry when we are unregistering the 1696 * device; see the function unregister_netdev 1697 * @close_list: List entry used when we are closing the device 1698 * @ptype_all: Device-specific packet handlers for all protocols 1699 * @ptype_specific: Device-specific, protocol-specific packet handlers 1700 * 1701 * @adj_list: Directly linked devices, like slaves for bonding 1702 * @features: Currently active device features 1703 * @hw_features: User-changeable features 1704 * 1705 * @wanted_features: User-requested features 1706 * @vlan_features: Mask of features inheritable by VLAN devices 1707 * 1708 * @hw_enc_features: Mask of features inherited by encapsulating devices 1709 * This field indicates what encapsulation 1710 * offloads the hardware is capable of doing, 1711 * and drivers will need to set them appropriately. 1712 * 1713 * @mpls_features: Mask of features inheritable by MPLS 1714 * @gso_partial_features: value(s) from NETIF_F_GSO\* 1715 * 1716 * @ifindex: interface index 1717 * @group: The group the device belongs to 1718 * 1719 * @stats: Statistics struct, which was left as a legacy, use 1720 * rtnl_link_stats64 instead 1721 * 1722 * @rx_dropped: Dropped packets by core network, 1723 * do not use this in drivers 1724 * @tx_dropped: Dropped packets by core network, 1725 * do not use this in drivers 1726 * @rx_nohandler: nohandler dropped packets by core network on 1727 * inactive devices, do not use this in drivers 1728 * @carrier_up_count: Number of times the carrier has been up 1729 * @carrier_down_count: Number of times the carrier has been down 1730 * 1731 * @wireless_handlers: List of functions to handle Wireless Extensions, 1732 * instead of ioctl, 1733 * see <net/iw_handler.h> for details. 1734 * @wireless_data: Instance data managed by the core of wireless extensions 1735 * 1736 * @netdev_ops: Includes several pointers to callbacks, 1737 * if one wants to override the ndo_*() functions 1738 * @ethtool_ops: Management operations 1739 * @l3mdev_ops: Layer 3 master device operations 1740 * @ndisc_ops: Includes callbacks for different IPv6 neighbour 1741 * discovery handling. Necessary for e.g. 6LoWPAN. 1742 * @xfrmdev_ops: Transformation offload operations 1743 * @tlsdev_ops: Transport Layer Security offload operations 1744 * @header_ops: Includes callbacks for creating,parsing,caching,etc 1745 * of Layer 2 headers. 1746 * 1747 * @flags: Interface flags (a la BSD) 1748 * @priv_flags: Like 'flags' but invisible to userspace, 1749 * see if.h for the definitions 1750 * @gflags: Global flags ( kept as legacy ) 1751 * @padded: How much padding added by alloc_netdev() 1752 * @operstate: RFC2863 operstate 1753 * @link_mode: Mapping policy to operstate 1754 * @if_port: Selectable AUI, TP, ... 1755 * @dma: DMA channel 1756 * @mtu: Interface MTU value 1757 * @min_mtu: Interface Minimum MTU value 1758 * @max_mtu: Interface Maximum MTU value 1759 * @type: Interface hardware type 1760 * @hard_header_len: Maximum hardware header length. 1761 * @min_header_len: Minimum hardware header length 1762 * 1763 * @needed_headroom: Extra headroom the hardware may need, but not in all 1764 * cases can this be guaranteed 1765 * @needed_tailroom: Extra tailroom the hardware may need, but not in all 1766 * cases can this be guaranteed. Some cases also use 1767 * LL_MAX_HEADER instead to allocate the skb 1768 * 1769 * interface address info: 1770 * 1771 * @perm_addr: Permanent hw address 1772 * @addr_assign_type: Hw address assignment type 1773 * @addr_len: Hardware address length 1774 * @upper_level: Maximum depth level of upper devices. 1775 * @lower_level: Maximum depth level of lower devices. 1776 * @neigh_priv_len: Used in neigh_alloc() 1777 * @dev_id: Used to differentiate devices that share 1778 * the same link layer address 1779 * @dev_port: Used to differentiate devices that share 1780 * the same function 1781 * @addr_list_lock: XXX: need comments on this one 1782 * @name_assign_type: network interface name assignment type 1783 * @uc_promisc: Counter that indicates promiscuous mode 1784 * has been enabled due to the need to listen to 1785 * additional unicast addresses in a device that 1786 * does not implement ndo_set_rx_mode() 1787 * @uc: unicast mac addresses 1788 * @mc: multicast mac addresses 1789 * @dev_addrs: list of device hw addresses 1790 * @queues_kset: Group of all Kobjects in the Tx and RX queues 1791 * @promiscuity: Number of times the NIC is told to work in 1792 * promiscuous mode; if it becomes 0 the NIC will 1793 * exit promiscuous mode 1794 * @allmulti: Counter, enables or disables allmulticast mode 1795 * 1796 * @vlan_info: VLAN info 1797 * @dsa_ptr: dsa specific data 1798 * @tipc_ptr: TIPC specific data 1799 * @atalk_ptr: AppleTalk link 1800 * @ip_ptr: IPv4 specific data 1801 * @dn_ptr: DECnet specific data 1802 * @ip6_ptr: IPv6 specific data 1803 * @ax25_ptr: AX.25 specific data 1804 * @ieee80211_ptr: IEEE 802.11 specific data, assign before registering 1805 * @ieee802154_ptr: IEEE 802.15.4 low-rate Wireless Personal Area Network 1806 * device struct 1807 * @mpls_ptr: mpls_dev struct pointer 1808 * 1809 * @dev_addr: Hw address (before bcast, 1810 * because most packets are unicast) 1811 * 1812 * @_rx: Array of RX queues 1813 * @num_rx_queues: Number of RX queues 1814 * allocated at register_netdev() time 1815 * @real_num_rx_queues: Number of RX queues currently active in device 1816 * @xdp_prog: XDP sockets filter program pointer 1817 * @gro_flush_timeout: timeout for GRO layer in NAPI 1818 * @napi_defer_hard_irqs: If not zero, provides a counter that would 1819 * allow to avoid NIC hard IRQ, on busy queues. 1820 * 1821 * @rx_handler: handler for received packets 1822 * @rx_handler_data: XXX: need comments on this one 1823 * @miniq_ingress: ingress/clsact qdisc specific data for 1824 * ingress processing 1825 * @ingress_queue: XXX: need comments on this one 1826 * @nf_hooks_ingress: netfilter hooks executed for ingress packets 1827 * @broadcast: hw bcast address 1828 * 1829 * @rx_cpu_rmap: CPU reverse-mapping for RX completion interrupts, 1830 * indexed by RX queue number. Assigned by driver. 1831 * This must only be set if the ndo_rx_flow_steer 1832 * operation is defined 1833 * @index_hlist: Device index hash chain 1834 * 1835 * @_tx: Array of TX queues 1836 * @num_tx_queues: Number of TX queues allocated at alloc_netdev_mq() time 1837 * @real_num_tx_queues: Number of TX queues currently active in device 1838 * @qdisc: Root qdisc from userspace point of view 1839 * @tx_queue_len: Max frames per queue allowed 1840 * @tx_global_lock: XXX: need comments on this one 1841 * @xdp_bulkq: XDP device bulk queue 1842 * @xps_maps: all CPUs/RXQs maps for XPS device 1843 * 1844 * @xps_maps: XXX: need comments on this one 1845 * @miniq_egress: clsact qdisc specific data for 1846 * egress processing 1847 * @qdisc_hash: qdisc hash table 1848 * @watchdog_timeo: Represents the timeout that is used by 1849 * the watchdog (see dev_watchdog()) 1850 * @watchdog_timer: List of timers 1851 * 1852 * @proto_down_reason: reason a netdev interface is held down 1853 * @pcpu_refcnt: Number of references to this device 1854 * @dev_refcnt: Number of references to this device 1855 * @todo_list: Delayed register/unregister 1856 * @link_watch_list: XXX: need comments on this one 1857 * 1858 * @reg_state: Register/unregister state machine 1859 * @dismantle: Device is going to be freed 1860 * @rtnl_link_state: This enum represents the phases of creating 1861 * a new link 1862 * 1863 * @needs_free_netdev: Should unregister perform free_netdev? 1864 * @priv_destructor: Called from unregister 1865 * @npinfo: XXX: need comments on this one 1866 * @nd_net: Network namespace this network device is inside 1867 * 1868 * @ml_priv: Mid-layer private 1869 * @ml_priv_type: Mid-layer private type 1870 * @lstats: Loopback statistics 1871 * @tstats: Tunnel statistics 1872 * @dstats: Dummy statistics 1873 * @vstats: Virtual ethernet statistics 1874 * 1875 * @garp_port: GARP 1876 * @mrp_port: MRP 1877 * 1878 * @dev: Class/net/name entry 1879 * @sysfs_groups: Space for optional device, statistics and wireless 1880 * sysfs groups 1881 * 1882 * @sysfs_rx_queue_group: Space for optional per-rx queue attributes 1883 * @rtnl_link_ops: Rtnl_link_ops 1884 * 1885 * @gso_max_size: Maximum size of generic segmentation offload 1886 * @gso_max_segs: Maximum number of segments that can be passed to the 1887 * NIC for GSO 1888 * 1889 * @dcbnl_ops: Data Center Bridging netlink ops 1890 * @num_tc: Number of traffic classes in the net device 1891 * @tc_to_txq: XXX: need comments on this one 1892 * @prio_tc_map: XXX: need comments on this one 1893 * 1894 * @fcoe_ddp_xid: Max exchange id for FCoE LRO by ddp 1895 * 1896 * @priomap: XXX: need comments on this one 1897 * @phydev: Physical device may attach itself 1898 * for hardware timestamping 1899 * @sfp_bus: attached &struct sfp_bus structure. 1900 * 1901 * @qdisc_tx_busylock: lockdep class annotating Qdisc->busylock spinlock 1902 * @qdisc_running_key: lockdep class annotating Qdisc->running seqcount 1903 * 1904 * @proto_down: protocol port state information can be sent to the 1905 * switch driver and used to set the phys state of the 1906 * switch port. 1907 * 1908 * @wol_enabled: Wake-on-LAN is enabled 1909 * 1910 * @threaded: napi threaded mode is enabled 1911 * 1912 * @net_notifier_list: List of per-net netdev notifier block 1913 * that follow this device when it is moved 1914 * to another network namespace. 1915 * 1916 * @macsec_ops: MACsec offloading ops 1917 * 1918 * @udp_tunnel_nic_info: static structure describing the UDP tunnel 1919 * offload capabilities of the device 1920 * @udp_tunnel_nic: UDP tunnel offload state 1921 * @xdp_state: stores info on attached XDP BPF programs 1922 * 1923 * @nested_level: Used as as a parameter of spin_lock_nested() of 1924 * dev->addr_list_lock. 1925 * @unlink_list: As netif_addr_lock() can be called recursively, 1926 * keep a list of interfaces to be deleted. 1927 * 1928 * FIXME: cleanup struct net_device such that network protocol info 1929 * moves out. 1930 */ 1931 1932 struct net_device { 1933 char name[IFNAMSIZ]; 1934 struct netdev_name_node *name_node; 1935 struct dev_ifalias __rcu *ifalias; 1936 /* 1937 * I/O specific fields 1938 * FIXME: Merge these and struct ifmap into one 1939 */ 1940 unsigned long mem_end; 1941 unsigned long mem_start; 1942 unsigned long base_addr; 1943 1944 /* 1945 * Some hardware also needs these fields (state,dev_list, 1946 * napi_list,unreg_list,close_list) but they are not 1947 * part of the usual set specified in Space.c. 1948 */ 1949 1950 unsigned long state; 1951 1952 struct list_head dev_list; 1953 struct list_head napi_list; 1954 struct list_head unreg_list; 1955 struct list_head close_list; 1956 struct list_head ptype_all; 1957 struct list_head ptype_specific; 1958 1959 struct { 1960 struct list_head upper; 1961 struct list_head lower; 1962 } adj_list; 1963 1964 /* Read-mostly cache-line for fast-path access */ 1965 unsigned int flags; 1966 unsigned int priv_flags; 1967 const struct net_device_ops *netdev_ops; 1968 int ifindex; 1969 unsigned short gflags; 1970 unsigned short hard_header_len; 1971 1972 /* Note : dev->mtu is often read without holding a lock. 1973 * Writers usually hold RTNL. 1974 * It is recommended to use READ_ONCE() to annotate the reads, 1975 * and to use WRITE_ONCE() to annotate the writes. 1976 */ 1977 unsigned int mtu; 1978 unsigned short needed_headroom; 1979 unsigned short needed_tailroom; 1980 1981 netdev_features_t features; 1982 netdev_features_t hw_features; 1983 netdev_features_t wanted_features; 1984 netdev_features_t vlan_features; 1985 netdev_features_t hw_enc_features; 1986 netdev_features_t mpls_features; 1987 netdev_features_t gso_partial_features; 1988 1989 unsigned int min_mtu; 1990 unsigned int max_mtu; 1991 unsigned short type; 1992 unsigned char min_header_len; 1993 unsigned char name_assign_type; 1994 1995 int group; 1996 1997 struct net_device_stats stats; /* not used by modern drivers */ 1998 1999 atomic_long_t rx_dropped; 2000 atomic_long_t tx_dropped; 2001 atomic_long_t rx_nohandler; 2002 2003 /* Stats to monitor link on/off, flapping */ 2004 atomic_t carrier_up_count; 2005 atomic_t carrier_down_count; 2006 2007 #ifdef CONFIG_WIRELESS_EXT 2008 const struct iw_handler_def *wireless_handlers; 2009 struct iw_public_data *wireless_data; 2010 #endif 2011 const struct ethtool_ops *ethtool_ops; 2012 #ifdef CONFIG_NET_L3_MASTER_DEV 2013 const struct l3mdev_ops *l3mdev_ops; 2014 #endif 2015 #if IS_ENABLED(CONFIG_IPV6) 2016 const struct ndisc_ops *ndisc_ops; 2017 #endif 2018 2019 #ifdef CONFIG_XFRM_OFFLOAD 2020 const struct xfrmdev_ops *xfrmdev_ops; 2021 #endif 2022 2023 #if IS_ENABLED(CONFIG_TLS_DEVICE) 2024 const struct tlsdev_ops *tlsdev_ops; 2025 #endif 2026 2027 const struct header_ops *header_ops; 2028 2029 unsigned char operstate; 2030 unsigned char link_mode; 2031 2032 unsigned char if_port; 2033 unsigned char dma; 2034 2035 /* Interface address info. */ 2036 unsigned char perm_addr[MAX_ADDR_LEN]; 2037 unsigned char addr_assign_type; 2038 unsigned char addr_len; 2039 unsigned char upper_level; 2040 unsigned char lower_level; 2041 2042 unsigned short neigh_priv_len; 2043 unsigned short dev_id; 2044 unsigned short dev_port; 2045 unsigned short padded; 2046 2047 spinlock_t addr_list_lock; 2048 int irq; 2049 2050 struct netdev_hw_addr_list uc; 2051 struct netdev_hw_addr_list mc; 2052 struct netdev_hw_addr_list dev_addrs; 2053 2054 #ifdef CONFIG_SYSFS 2055 struct kset *queues_kset; 2056 #endif 2057 #ifdef CONFIG_LOCKDEP 2058 struct list_head unlink_list; 2059 #endif 2060 unsigned int promiscuity; 2061 unsigned int allmulti; 2062 bool uc_promisc; 2063 #ifdef CONFIG_LOCKDEP 2064 unsigned char nested_level; 2065 #endif 2066 2067 2068 /* Protocol-specific pointers */ 2069 2070 #if IS_ENABLED(CONFIG_VLAN_8021Q) 2071 struct vlan_info __rcu *vlan_info; 2072 #endif 2073 #if IS_ENABLED(CONFIG_NET_DSA) 2074 struct dsa_port *dsa_ptr; 2075 #endif 2076 #if IS_ENABLED(CONFIG_TIPC) 2077 struct tipc_bearer __rcu *tipc_ptr; 2078 #endif 2079 #if IS_ENABLED(CONFIG_IRDA) || IS_ENABLED(CONFIG_ATALK) 2080 void *atalk_ptr; 2081 #endif 2082 struct in_device __rcu *ip_ptr; 2083 #if IS_ENABLED(CONFIG_DECNET) 2084 struct dn_dev __rcu *dn_ptr; 2085 #endif 2086 struct inet6_dev __rcu *ip6_ptr; 2087 #if IS_ENABLED(CONFIG_AX25) 2088 void *ax25_ptr; 2089 #endif 2090 struct wireless_dev *ieee80211_ptr; 2091 struct wpan_dev *ieee802154_ptr; 2092 #if IS_ENABLED(CONFIG_MPLS_ROUTING) 2093 struct mpls_dev __rcu *mpls_ptr; 2094 #endif 2095 2096 /* 2097 * Cache lines mostly used on receive path (including eth_type_trans()) 2098 */ 2099 /* Interface address info used in eth_type_trans() */ 2100 unsigned char *dev_addr; 2101 2102 struct netdev_rx_queue *_rx; 2103 unsigned int num_rx_queues; 2104 unsigned int real_num_rx_queues; 2105 2106 struct bpf_prog __rcu *xdp_prog; 2107 unsigned long gro_flush_timeout; 2108 int napi_defer_hard_irqs; 2109 rx_handler_func_t __rcu *rx_handler; 2110 void __rcu *rx_handler_data; 2111 2112 #ifdef CONFIG_NET_CLS_ACT 2113 struct mini_Qdisc __rcu *miniq_ingress; 2114 #endif 2115 struct netdev_queue __rcu *ingress_queue; 2116 #ifdef CONFIG_NETFILTER_INGRESS 2117 struct nf_hook_entries __rcu *nf_hooks_ingress; 2118 #endif 2119 2120 unsigned char broadcast[MAX_ADDR_LEN]; 2121 #ifdef CONFIG_RFS_ACCEL 2122 struct cpu_rmap *rx_cpu_rmap; 2123 #endif 2124 struct hlist_node index_hlist; 2125 2126 /* 2127 * Cache lines mostly used on transmit path 2128 */ 2129 struct netdev_queue *_tx ____cacheline_aligned_in_smp; 2130 unsigned int num_tx_queues; 2131 unsigned int real_num_tx_queues; 2132 struct Qdisc *qdisc; 2133 unsigned int tx_queue_len; 2134 spinlock_t tx_global_lock; 2135 2136 struct xdp_dev_bulk_queue __percpu *xdp_bulkq; 2137 2138 #ifdef CONFIG_XPS 2139 struct xps_dev_maps __rcu *xps_maps[XPS_MAPS_MAX]; 2140 #endif 2141 #ifdef CONFIG_NET_CLS_ACT 2142 struct mini_Qdisc __rcu *miniq_egress; 2143 #endif 2144 2145 #ifdef CONFIG_NET_SCHED 2146 DECLARE_HASHTABLE (qdisc_hash, 4); 2147 #endif 2148 /* These may be needed for future network-power-down code. */ 2149 struct timer_list watchdog_timer; 2150 int watchdog_timeo; 2151 2152 u32 proto_down_reason; 2153 2154 struct list_head todo_list; 2155 2156 #ifdef CONFIG_PCPU_DEV_REFCNT 2157 int __percpu *pcpu_refcnt; 2158 #else 2159 refcount_t dev_refcnt; 2160 #endif 2161 2162 struct list_head link_watch_list; 2163 2164 enum { NETREG_UNINITIALIZED=0, 2165 NETREG_REGISTERED, /* completed register_netdevice */ 2166 NETREG_UNREGISTERING, /* called unregister_netdevice */ 2167 NETREG_UNREGISTERED, /* completed unregister todo */ 2168 NETREG_RELEASED, /* called free_netdev */ 2169 NETREG_DUMMY, /* dummy device for NAPI poll */ 2170 } reg_state:8; 2171 2172 bool dismantle; 2173 2174 enum { 2175 RTNL_LINK_INITIALIZED, 2176 RTNL_LINK_INITIALIZING, 2177 } rtnl_link_state:16; 2178 2179 bool needs_free_netdev; 2180 void (*priv_destructor)(struct net_device *dev); 2181 2182 #ifdef CONFIG_NETPOLL 2183 struct netpoll_info __rcu *npinfo; 2184 #endif 2185 2186 possible_net_t nd_net; 2187 2188 /* mid-layer private */ 2189 void *ml_priv; 2190 enum netdev_ml_priv_type ml_priv_type; 2191 2192 union { 2193 struct pcpu_lstats __percpu *lstats; 2194 struct pcpu_sw_netstats __percpu *tstats; 2195 struct pcpu_dstats __percpu *dstats; 2196 }; 2197 2198 #if IS_ENABLED(CONFIG_GARP) 2199 struct garp_port __rcu *garp_port; 2200 #endif 2201 #if IS_ENABLED(CONFIG_MRP) 2202 struct mrp_port __rcu *mrp_port; 2203 #endif 2204 2205 struct device dev; 2206 const struct attribute_group *sysfs_groups[4]; 2207 const struct attribute_group *sysfs_rx_queue_group; 2208 2209 const struct rtnl_link_ops *rtnl_link_ops; 2210 2211 /* for setting kernel sock attribute on TCP connection setup */ 2212 #define GSO_MAX_SIZE 65536 2213 unsigned int gso_max_size; 2214 #define GSO_MAX_SEGS 65535 2215 u16 gso_max_segs; 2216 2217 #ifdef CONFIG_DCB 2218 const struct dcbnl_rtnl_ops *dcbnl_ops; 2219 #endif 2220 s16 num_tc; 2221 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE]; 2222 u8 prio_tc_map[TC_BITMASK + 1]; 2223 2224 #if IS_ENABLED(CONFIG_FCOE) 2225 unsigned int fcoe_ddp_xid; 2226 #endif 2227 #if IS_ENABLED(CONFIG_CGROUP_NET_PRIO) 2228 struct netprio_map __rcu *priomap; 2229 #endif 2230 struct phy_device *phydev; 2231 struct sfp_bus *sfp_bus; 2232 struct lock_class_key *qdisc_tx_busylock; 2233 struct lock_class_key *qdisc_running_key; 2234 bool proto_down; 2235 unsigned wol_enabled:1; 2236 unsigned threaded:1; 2237 2238 struct list_head net_notifier_list; 2239 2240 #if IS_ENABLED(CONFIG_MACSEC) 2241 /* MACsec management functions */ 2242 const struct macsec_ops *macsec_ops; 2243 #endif 2244 const struct udp_tunnel_nic_info *udp_tunnel_nic_info; 2245 struct udp_tunnel_nic *udp_tunnel_nic; 2246 2247 /* protected by rtnl_lock */ 2248 struct bpf_xdp_entity xdp_state[__MAX_XDP_MODE]; 2249 }; 2250 #define to_net_dev(d) container_of(d, struct net_device, dev) 2251 2252 static inline bool netif_elide_gro(const struct net_device *dev) 2253 { 2254 if (!(dev->features & NETIF_F_GRO) || dev->xdp_prog) 2255 return true; 2256 return false; 2257 } 2258 2259 #define NETDEV_ALIGN 32 2260 2261 static inline 2262 int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio) 2263 { 2264 return dev->prio_tc_map[prio & TC_BITMASK]; 2265 } 2266 2267 static inline 2268 int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc) 2269 { 2270 if (tc >= dev->num_tc) 2271 return -EINVAL; 2272 2273 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK; 2274 return 0; 2275 } 2276 2277 int netdev_txq_to_tc(struct net_device *dev, unsigned int txq); 2278 void netdev_reset_tc(struct net_device *dev); 2279 int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset); 2280 int netdev_set_num_tc(struct net_device *dev, u8 num_tc); 2281 2282 static inline 2283 int netdev_get_num_tc(struct net_device *dev) 2284 { 2285 return dev->num_tc; 2286 } 2287 2288 static inline void net_prefetch(void *p) 2289 { 2290 prefetch(p); 2291 #if L1_CACHE_BYTES < 128 2292 prefetch((u8 *)p + L1_CACHE_BYTES); 2293 #endif 2294 } 2295 2296 static inline void net_prefetchw(void *p) 2297 { 2298 prefetchw(p); 2299 #if L1_CACHE_BYTES < 128 2300 prefetchw((u8 *)p + L1_CACHE_BYTES); 2301 #endif 2302 } 2303 2304 void netdev_unbind_sb_channel(struct net_device *dev, 2305 struct net_device *sb_dev); 2306 int netdev_bind_sb_channel_queue(struct net_device *dev, 2307 struct net_device *sb_dev, 2308 u8 tc, u16 count, u16 offset); 2309 int netdev_set_sb_channel(struct net_device *dev, u16 channel); 2310 static inline int netdev_get_sb_channel(struct net_device *dev) 2311 { 2312 return max_t(int, -dev->num_tc, 0); 2313 } 2314 2315 static inline 2316 struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev, 2317 unsigned int index) 2318 { 2319 return &dev->_tx[index]; 2320 } 2321 2322 static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev, 2323 const struct sk_buff *skb) 2324 { 2325 return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb)); 2326 } 2327 2328 static inline void netdev_for_each_tx_queue(struct net_device *dev, 2329 void (*f)(struct net_device *, 2330 struct netdev_queue *, 2331 void *), 2332 void *arg) 2333 { 2334 unsigned int i; 2335 2336 for (i = 0; i < dev->num_tx_queues; i++) 2337 f(dev, &dev->_tx[i], arg); 2338 } 2339 2340 #define netdev_lockdep_set_classes(dev) \ 2341 { \ 2342 static struct lock_class_key qdisc_tx_busylock_key; \ 2343 static struct lock_class_key qdisc_running_key; \ 2344 static struct lock_class_key qdisc_xmit_lock_key; \ 2345 static struct lock_class_key dev_addr_list_lock_key; \ 2346 unsigned int i; \ 2347 \ 2348 (dev)->qdisc_tx_busylock = &qdisc_tx_busylock_key; \ 2349 (dev)->qdisc_running_key = &qdisc_running_key; \ 2350 lockdep_set_class(&(dev)->addr_list_lock, \ 2351 &dev_addr_list_lock_key); \ 2352 for (i = 0; i < (dev)->num_tx_queues; i++) \ 2353 lockdep_set_class(&(dev)->_tx[i]._xmit_lock, \ 2354 &qdisc_xmit_lock_key); \ 2355 } 2356 2357 u16 netdev_pick_tx(struct net_device *dev, struct sk_buff *skb, 2358 struct net_device *sb_dev); 2359 struct netdev_queue *netdev_core_pick_tx(struct net_device *dev, 2360 struct sk_buff *skb, 2361 struct net_device *sb_dev); 2362 2363 /* returns the headroom that the master device needs to take in account 2364 * when forwarding to this dev 2365 */ 2366 static inline unsigned netdev_get_fwd_headroom(struct net_device *dev) 2367 { 2368 return dev->priv_flags & IFF_PHONY_HEADROOM ? 0 : dev->needed_headroom; 2369 } 2370 2371 static inline void netdev_set_rx_headroom(struct net_device *dev, int new_hr) 2372 { 2373 if (dev->netdev_ops->ndo_set_rx_headroom) 2374 dev->netdev_ops->ndo_set_rx_headroom(dev, new_hr); 2375 } 2376 2377 /* set the device rx headroom to the dev's default */ 2378 static inline void netdev_reset_rx_headroom(struct net_device *dev) 2379 { 2380 netdev_set_rx_headroom(dev, -1); 2381 } 2382 2383 static inline void *netdev_get_ml_priv(struct net_device *dev, 2384 enum netdev_ml_priv_type type) 2385 { 2386 if (dev->ml_priv_type != type) 2387 return NULL; 2388 2389 return dev->ml_priv; 2390 } 2391 2392 static inline void netdev_set_ml_priv(struct net_device *dev, 2393 void *ml_priv, 2394 enum netdev_ml_priv_type type) 2395 { 2396 WARN(dev->ml_priv_type && dev->ml_priv_type != type, 2397 "Overwriting already set ml_priv_type (%u) with different ml_priv_type (%u)!\n", 2398 dev->ml_priv_type, type); 2399 WARN(!dev->ml_priv_type && dev->ml_priv, 2400 "Overwriting already set ml_priv and ml_priv_type is ML_PRIV_NONE!\n"); 2401 2402 dev->ml_priv = ml_priv; 2403 dev->ml_priv_type = type; 2404 } 2405 2406 /* 2407 * Net namespace inlines 2408 */ 2409 static inline 2410 struct net *dev_net(const struct net_device *dev) 2411 { 2412 return read_pnet(&dev->nd_net); 2413 } 2414 2415 static inline 2416 void dev_net_set(struct net_device *dev, struct net *net) 2417 { 2418 write_pnet(&dev->nd_net, net); 2419 } 2420 2421 /** 2422 * netdev_priv - access network device private data 2423 * @dev: network device 2424 * 2425 * Get network device private data 2426 */ 2427 static inline void *netdev_priv(const struct net_device *dev) 2428 { 2429 return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN); 2430 } 2431 2432 /* Set the sysfs physical device reference for the network logical device 2433 * if set prior to registration will cause a symlink during initialization. 2434 */ 2435 #define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev)) 2436 2437 /* Set the sysfs device type for the network logical device to allow 2438 * fine-grained identification of different network device types. For 2439 * example Ethernet, Wireless LAN, Bluetooth, WiMAX etc. 2440 */ 2441 #define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype)) 2442 2443 /* Default NAPI poll() weight 2444 * Device drivers are strongly advised to not use bigger value 2445 */ 2446 #define NAPI_POLL_WEIGHT 64 2447 2448 /** 2449 * netif_napi_add - initialize a NAPI context 2450 * @dev: network device 2451 * @napi: NAPI context 2452 * @poll: polling function 2453 * @weight: default weight 2454 * 2455 * netif_napi_add() must be used to initialize a NAPI context prior to calling 2456 * *any* of the other NAPI-related functions. 2457 */ 2458 void netif_napi_add(struct net_device *dev, struct napi_struct *napi, 2459 int (*poll)(struct napi_struct *, int), int weight); 2460 2461 /** 2462 * netif_tx_napi_add - initialize a NAPI context 2463 * @dev: network device 2464 * @napi: NAPI context 2465 * @poll: polling function 2466 * @weight: default weight 2467 * 2468 * This variant of netif_napi_add() should be used from drivers using NAPI 2469 * to exclusively poll a TX queue. 2470 * This will avoid we add it into napi_hash[], thus polluting this hash table. 2471 */ 2472 static inline void netif_tx_napi_add(struct net_device *dev, 2473 struct napi_struct *napi, 2474 int (*poll)(struct napi_struct *, int), 2475 int weight) 2476 { 2477 set_bit(NAPI_STATE_NO_BUSY_POLL, &napi->state); 2478 netif_napi_add(dev, napi, poll, weight); 2479 } 2480 2481 /** 2482 * __netif_napi_del - remove a NAPI context 2483 * @napi: NAPI context 2484 * 2485 * Warning: caller must observe RCU grace period before freeing memory 2486 * containing @napi. Drivers might want to call this helper to combine 2487 * all the needed RCU grace periods into a single one. 2488 */ 2489 void __netif_napi_del(struct napi_struct *napi); 2490 2491 /** 2492 * netif_napi_del - remove a NAPI context 2493 * @napi: NAPI context 2494 * 2495 * netif_napi_del() removes a NAPI context from the network device NAPI list 2496 */ 2497 static inline void netif_napi_del(struct napi_struct *napi) 2498 { 2499 __netif_napi_del(napi); 2500 synchronize_net(); 2501 } 2502 2503 struct napi_gro_cb { 2504 /* Virtual address of skb_shinfo(skb)->frags[0].page + offset. */ 2505 void *frag0; 2506 2507 /* Length of frag0. */ 2508 unsigned int frag0_len; 2509 2510 /* This indicates where we are processing relative to skb->data. */ 2511 int data_offset; 2512 2513 /* This is non-zero if the packet cannot be merged with the new skb. */ 2514 u16 flush; 2515 2516 /* Save the IP ID here and check when we get to the transport layer */ 2517 u16 flush_id; 2518 2519 /* Number of segments aggregated. */ 2520 u16 count; 2521 2522 /* Start offset for remote checksum offload */ 2523 u16 gro_remcsum_start; 2524 2525 /* jiffies when first packet was created/queued */ 2526 unsigned long age; 2527 2528 /* Used in ipv6_gro_receive() and foo-over-udp */ 2529 u16 proto; 2530 2531 /* This is non-zero if the packet may be of the same flow. */ 2532 u8 same_flow:1; 2533 2534 /* Used in tunnel GRO receive */ 2535 u8 encap_mark:1; 2536 2537 /* GRO checksum is valid */ 2538 u8 csum_valid:1; 2539 2540 /* Number of checksums via CHECKSUM_UNNECESSARY */ 2541 u8 csum_cnt:3; 2542 2543 /* Free the skb? */ 2544 u8 free:2; 2545 #define NAPI_GRO_FREE 1 2546 #define NAPI_GRO_FREE_STOLEN_HEAD 2 2547 2548 /* Used in foo-over-udp, set in udp[46]_gro_receive */ 2549 u8 is_ipv6:1; 2550 2551 /* Used in GRE, set in fou/gue_gro_receive */ 2552 u8 is_fou:1; 2553 2554 /* Used to determine if flush_id can be ignored */ 2555 u8 is_atomic:1; 2556 2557 /* Number of gro_receive callbacks this packet already went through */ 2558 u8 recursion_counter:4; 2559 2560 /* GRO is done by frag_list pointer chaining. */ 2561 u8 is_flist:1; 2562 2563 /* used to support CHECKSUM_COMPLETE for tunneling protocols */ 2564 __wsum csum; 2565 2566 /* used in skb_gro_receive() slow path */ 2567 struct sk_buff *last; 2568 }; 2569 2570 #define NAPI_GRO_CB(skb) ((struct napi_gro_cb *)(skb)->cb) 2571 2572 #define GRO_RECURSION_LIMIT 15 2573 static inline int gro_recursion_inc_test(struct sk_buff *skb) 2574 { 2575 return ++NAPI_GRO_CB(skb)->recursion_counter == GRO_RECURSION_LIMIT; 2576 } 2577 2578 typedef struct sk_buff *(*gro_receive_t)(struct list_head *, struct sk_buff *); 2579 static inline struct sk_buff *call_gro_receive(gro_receive_t cb, 2580 struct list_head *head, 2581 struct sk_buff *skb) 2582 { 2583 if (unlikely(gro_recursion_inc_test(skb))) { 2584 NAPI_GRO_CB(skb)->flush |= 1; 2585 return NULL; 2586 } 2587 2588 return cb(head, skb); 2589 } 2590 2591 typedef struct sk_buff *(*gro_receive_sk_t)(struct sock *, struct list_head *, 2592 struct sk_buff *); 2593 static inline struct sk_buff *call_gro_receive_sk(gro_receive_sk_t cb, 2594 struct sock *sk, 2595 struct list_head *head, 2596 struct sk_buff *skb) 2597 { 2598 if (unlikely(gro_recursion_inc_test(skb))) { 2599 NAPI_GRO_CB(skb)->flush |= 1; 2600 return NULL; 2601 } 2602 2603 return cb(sk, head, skb); 2604 } 2605 2606 struct packet_type { 2607 __be16 type; /* This is really htons(ether_type). */ 2608 bool ignore_outgoing; 2609 struct net_device *dev; /* NULL is wildcarded here */ 2610 int (*func) (struct sk_buff *, 2611 struct net_device *, 2612 struct packet_type *, 2613 struct net_device *); 2614 void (*list_func) (struct list_head *, 2615 struct packet_type *, 2616 struct net_device *); 2617 bool (*id_match)(struct packet_type *ptype, 2618 struct sock *sk); 2619 void *af_packet_priv; 2620 struct list_head list; 2621 }; 2622 2623 struct offload_callbacks { 2624 struct sk_buff *(*gso_segment)(struct sk_buff *skb, 2625 netdev_features_t features); 2626 struct sk_buff *(*gro_receive)(struct list_head *head, 2627 struct sk_buff *skb); 2628 int (*gro_complete)(struct sk_buff *skb, int nhoff); 2629 }; 2630 2631 struct packet_offload { 2632 __be16 type; /* This is really htons(ether_type). */ 2633 u16 priority; 2634 struct offload_callbacks callbacks; 2635 struct list_head list; 2636 }; 2637 2638 /* often modified stats are per-CPU, other are shared (netdev->stats) */ 2639 struct pcpu_sw_netstats { 2640 u64 rx_packets; 2641 u64 rx_bytes; 2642 u64 tx_packets; 2643 u64 tx_bytes; 2644 struct u64_stats_sync syncp; 2645 } __aligned(4 * sizeof(u64)); 2646 2647 struct pcpu_lstats { 2648 u64_stats_t packets; 2649 u64_stats_t bytes; 2650 struct u64_stats_sync syncp; 2651 } __aligned(2 * sizeof(u64)); 2652 2653 void dev_lstats_read(struct net_device *dev, u64 *packets, u64 *bytes); 2654 2655 static inline void dev_sw_netstats_rx_add(struct net_device *dev, unsigned int len) 2656 { 2657 struct pcpu_sw_netstats *tstats = this_cpu_ptr(dev->tstats); 2658 2659 u64_stats_update_begin(&tstats->syncp); 2660 tstats->rx_bytes += len; 2661 tstats->rx_packets++; 2662 u64_stats_update_end(&tstats->syncp); 2663 } 2664 2665 static inline void dev_sw_netstats_tx_add(struct net_device *dev, 2666 unsigned int packets, 2667 unsigned int len) 2668 { 2669 struct pcpu_sw_netstats *tstats = this_cpu_ptr(dev->tstats); 2670 2671 u64_stats_update_begin(&tstats->syncp); 2672 tstats->tx_bytes += len; 2673 tstats->tx_packets += packets; 2674 u64_stats_update_end(&tstats->syncp); 2675 } 2676 2677 static inline void dev_lstats_add(struct net_device *dev, unsigned int len) 2678 { 2679 struct pcpu_lstats *lstats = this_cpu_ptr(dev->lstats); 2680 2681 u64_stats_update_begin(&lstats->syncp); 2682 u64_stats_add(&lstats->bytes, len); 2683 u64_stats_inc(&lstats->packets); 2684 u64_stats_update_end(&lstats->syncp); 2685 } 2686 2687 #define __netdev_alloc_pcpu_stats(type, gfp) \ 2688 ({ \ 2689 typeof(type) __percpu *pcpu_stats = alloc_percpu_gfp(type, gfp);\ 2690 if (pcpu_stats) { \ 2691 int __cpu; \ 2692 for_each_possible_cpu(__cpu) { \ 2693 typeof(type) *stat; \ 2694 stat = per_cpu_ptr(pcpu_stats, __cpu); \ 2695 u64_stats_init(&stat->syncp); \ 2696 } \ 2697 } \ 2698 pcpu_stats; \ 2699 }) 2700 2701 #define netdev_alloc_pcpu_stats(type) \ 2702 __netdev_alloc_pcpu_stats(type, GFP_KERNEL) 2703 2704 #define devm_netdev_alloc_pcpu_stats(dev, type) \ 2705 ({ \ 2706 typeof(type) __percpu *pcpu_stats = devm_alloc_percpu(dev, type);\ 2707 if (pcpu_stats) { \ 2708 int __cpu; \ 2709 for_each_possible_cpu(__cpu) { \ 2710 typeof(type) *stat; \ 2711 stat = per_cpu_ptr(pcpu_stats, __cpu); \ 2712 u64_stats_init(&stat->syncp); \ 2713 } \ 2714 } \ 2715 pcpu_stats; \ 2716 }) 2717 2718 enum netdev_lag_tx_type { 2719 NETDEV_LAG_TX_TYPE_UNKNOWN, 2720 NETDEV_LAG_TX_TYPE_RANDOM, 2721 NETDEV_LAG_TX_TYPE_BROADCAST, 2722 NETDEV_LAG_TX_TYPE_ROUNDROBIN, 2723 NETDEV_LAG_TX_TYPE_ACTIVEBACKUP, 2724 NETDEV_LAG_TX_TYPE_HASH, 2725 }; 2726 2727 enum netdev_lag_hash { 2728 NETDEV_LAG_HASH_NONE, 2729 NETDEV_LAG_HASH_L2, 2730 NETDEV_LAG_HASH_L34, 2731 NETDEV_LAG_HASH_L23, 2732 NETDEV_LAG_HASH_E23, 2733 NETDEV_LAG_HASH_E34, 2734 NETDEV_LAG_HASH_VLAN_SRCMAC, 2735 NETDEV_LAG_HASH_UNKNOWN, 2736 }; 2737 2738 struct netdev_lag_upper_info { 2739 enum netdev_lag_tx_type tx_type; 2740 enum netdev_lag_hash hash_type; 2741 }; 2742 2743 struct netdev_lag_lower_state_info { 2744 u8 link_up : 1, 2745 tx_enabled : 1; 2746 }; 2747 2748 #include <linux/notifier.h> 2749 2750 /* netdevice notifier chain. Please remember to update netdev_cmd_to_name() 2751 * and the rtnetlink notification exclusion list in rtnetlink_event() when 2752 * adding new types. 2753 */ 2754 enum netdev_cmd { 2755 NETDEV_UP = 1, /* For now you can't veto a device up/down */ 2756 NETDEV_DOWN, 2757 NETDEV_REBOOT, /* Tell a protocol stack a network interface 2758 detected a hardware crash and restarted 2759 - we can use this eg to kick tcp sessions 2760 once done */ 2761 NETDEV_CHANGE, /* Notify device state change */ 2762 NETDEV_REGISTER, 2763 NETDEV_UNREGISTER, 2764 NETDEV_CHANGEMTU, /* notify after mtu change happened */ 2765 NETDEV_CHANGEADDR, /* notify after the address change */ 2766 NETDEV_PRE_CHANGEADDR, /* notify before the address change */ 2767 NETDEV_GOING_DOWN, 2768 NETDEV_CHANGENAME, 2769 NETDEV_FEAT_CHANGE, 2770 NETDEV_BONDING_FAILOVER, 2771 NETDEV_PRE_UP, 2772 NETDEV_PRE_TYPE_CHANGE, 2773 NETDEV_POST_TYPE_CHANGE, 2774 NETDEV_POST_INIT, 2775 NETDEV_RELEASE, 2776 NETDEV_NOTIFY_PEERS, 2777 NETDEV_JOIN, 2778 NETDEV_CHANGEUPPER, 2779 NETDEV_RESEND_IGMP, 2780 NETDEV_PRECHANGEMTU, /* notify before mtu change happened */ 2781 NETDEV_CHANGEINFODATA, 2782 NETDEV_BONDING_INFO, 2783 NETDEV_PRECHANGEUPPER, 2784 NETDEV_CHANGELOWERSTATE, 2785 NETDEV_UDP_TUNNEL_PUSH_INFO, 2786 NETDEV_UDP_TUNNEL_DROP_INFO, 2787 NETDEV_CHANGE_TX_QUEUE_LEN, 2788 NETDEV_CVLAN_FILTER_PUSH_INFO, 2789 NETDEV_CVLAN_FILTER_DROP_INFO, 2790 NETDEV_SVLAN_FILTER_PUSH_INFO, 2791 NETDEV_SVLAN_FILTER_DROP_INFO, 2792 }; 2793 const char *netdev_cmd_to_name(enum netdev_cmd cmd); 2794 2795 int register_netdevice_notifier(struct notifier_block *nb); 2796 int unregister_netdevice_notifier(struct notifier_block *nb); 2797 int register_netdevice_notifier_net(struct net *net, struct notifier_block *nb); 2798 int unregister_netdevice_notifier_net(struct net *net, 2799 struct notifier_block *nb); 2800 int register_netdevice_notifier_dev_net(struct net_device *dev, 2801 struct notifier_block *nb, 2802 struct netdev_net_notifier *nn); 2803 int unregister_netdevice_notifier_dev_net(struct net_device *dev, 2804 struct notifier_block *nb, 2805 struct netdev_net_notifier *nn); 2806 2807 struct netdev_notifier_info { 2808 struct net_device *dev; 2809 struct netlink_ext_ack *extack; 2810 }; 2811 2812 struct netdev_notifier_info_ext { 2813 struct netdev_notifier_info info; /* must be first */ 2814 union { 2815 u32 mtu; 2816 } ext; 2817 }; 2818 2819 struct netdev_notifier_change_info { 2820 struct netdev_notifier_info info; /* must be first */ 2821 unsigned int flags_changed; 2822 }; 2823 2824 struct netdev_notifier_changeupper_info { 2825 struct netdev_notifier_info info; /* must be first */ 2826 struct net_device *upper_dev; /* new upper dev */ 2827 bool master; /* is upper dev master */ 2828 bool linking; /* is the notification for link or unlink */ 2829 void *upper_info; /* upper dev info */ 2830 }; 2831 2832 struct netdev_notifier_changelowerstate_info { 2833 struct netdev_notifier_info info; /* must be first */ 2834 void *lower_state_info; /* is lower dev state */ 2835 }; 2836 2837 struct netdev_notifier_pre_changeaddr_info { 2838 struct netdev_notifier_info info; /* must be first */ 2839 const unsigned char *dev_addr; 2840 }; 2841 2842 static inline void netdev_notifier_info_init(struct netdev_notifier_info *info, 2843 struct net_device *dev) 2844 { 2845 info->dev = dev; 2846 info->extack = NULL; 2847 } 2848 2849 static inline struct net_device * 2850 netdev_notifier_info_to_dev(const struct netdev_notifier_info *info) 2851 { 2852 return info->dev; 2853 } 2854 2855 static inline struct netlink_ext_ack * 2856 netdev_notifier_info_to_extack(const struct netdev_notifier_info *info) 2857 { 2858 return info->extack; 2859 } 2860 2861 int call_netdevice_notifiers(unsigned long val, struct net_device *dev); 2862 2863 2864 extern rwlock_t dev_base_lock; /* Device list lock */ 2865 2866 #define for_each_netdev(net, d) \ 2867 list_for_each_entry(d, &(net)->dev_base_head, dev_list) 2868 #define for_each_netdev_reverse(net, d) \ 2869 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list) 2870 #define for_each_netdev_rcu(net, d) \ 2871 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list) 2872 #define for_each_netdev_safe(net, d, n) \ 2873 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list) 2874 #define for_each_netdev_continue(net, d) \ 2875 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list) 2876 #define for_each_netdev_continue_reverse(net, d) \ 2877 list_for_each_entry_continue_reverse(d, &(net)->dev_base_head, \ 2878 dev_list) 2879 #define for_each_netdev_continue_rcu(net, d) \ 2880 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list) 2881 #define for_each_netdev_in_bond_rcu(bond, slave) \ 2882 for_each_netdev_rcu(&init_net, slave) \ 2883 if (netdev_master_upper_dev_get_rcu(slave) == (bond)) 2884 #define net_device_entry(lh) list_entry(lh, struct net_device, dev_list) 2885 2886 static inline struct net_device *next_net_device(struct net_device *dev) 2887 { 2888 struct list_head *lh; 2889 struct net *net; 2890 2891 net = dev_net(dev); 2892 lh = dev->dev_list.next; 2893 return lh == &net->dev_base_head ? NULL : net_device_entry(lh); 2894 } 2895 2896 static inline struct net_device *next_net_device_rcu(struct net_device *dev) 2897 { 2898 struct list_head *lh; 2899 struct net *net; 2900 2901 net = dev_net(dev); 2902 lh = rcu_dereference(list_next_rcu(&dev->dev_list)); 2903 return lh == &net->dev_base_head ? NULL : net_device_entry(lh); 2904 } 2905 2906 static inline struct net_device *first_net_device(struct net *net) 2907 { 2908 return list_empty(&net->dev_base_head) ? NULL : 2909 net_device_entry(net->dev_base_head.next); 2910 } 2911 2912 static inline struct net_device *first_net_device_rcu(struct net *net) 2913 { 2914 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head)); 2915 2916 return lh == &net->dev_base_head ? NULL : net_device_entry(lh); 2917 } 2918 2919 int netdev_boot_setup_check(struct net_device *dev); 2920 unsigned long netdev_boot_base(const char *prefix, int unit); 2921 struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type, 2922 const char *hwaddr); 2923 struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type); 2924 void dev_add_pack(struct packet_type *pt); 2925 void dev_remove_pack(struct packet_type *pt); 2926 void __dev_remove_pack(struct packet_type *pt); 2927 void dev_add_offload(struct packet_offload *po); 2928 void dev_remove_offload(struct packet_offload *po); 2929 2930 int dev_get_iflink(const struct net_device *dev); 2931 int dev_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb); 2932 int dev_fill_forward_path(const struct net_device *dev, const u8 *daddr, 2933 struct net_device_path_stack *stack); 2934 struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags, 2935 unsigned short mask); 2936 struct net_device *dev_get_by_name(struct net *net, const char *name); 2937 struct net_device *dev_get_by_name_rcu(struct net *net, const char *name); 2938 struct net_device *__dev_get_by_name(struct net *net, const char *name); 2939 int dev_alloc_name(struct net_device *dev, const char *name); 2940 int dev_open(struct net_device *dev, struct netlink_ext_ack *extack); 2941 void dev_close(struct net_device *dev); 2942 void dev_close_many(struct list_head *head, bool unlink); 2943 void dev_disable_lro(struct net_device *dev); 2944 int dev_loopback_xmit(struct net *net, struct sock *sk, struct sk_buff *newskb); 2945 u16 dev_pick_tx_zero(struct net_device *dev, struct sk_buff *skb, 2946 struct net_device *sb_dev); 2947 u16 dev_pick_tx_cpu_id(struct net_device *dev, struct sk_buff *skb, 2948 struct net_device *sb_dev); 2949 2950 int dev_queue_xmit(struct sk_buff *skb); 2951 int dev_queue_xmit_accel(struct sk_buff *skb, struct net_device *sb_dev); 2952 int __dev_direct_xmit(struct sk_buff *skb, u16 queue_id); 2953 2954 static inline int dev_direct_xmit(struct sk_buff *skb, u16 queue_id) 2955 { 2956 int ret; 2957 2958 ret = __dev_direct_xmit(skb, queue_id); 2959 if (!dev_xmit_complete(ret)) 2960 kfree_skb(skb); 2961 return ret; 2962 } 2963 2964 int register_netdevice(struct net_device *dev); 2965 void unregister_netdevice_queue(struct net_device *dev, struct list_head *head); 2966 void unregister_netdevice_many(struct list_head *head); 2967 static inline void unregister_netdevice(struct net_device *dev) 2968 { 2969 unregister_netdevice_queue(dev, NULL); 2970 } 2971 2972 int netdev_refcnt_read(const struct net_device *dev); 2973 void free_netdev(struct net_device *dev); 2974 void netdev_freemem(struct net_device *dev); 2975 int init_dummy_netdev(struct net_device *dev); 2976 2977 struct net_device *netdev_get_xmit_slave(struct net_device *dev, 2978 struct sk_buff *skb, 2979 bool all_slaves); 2980 struct net_device *netdev_sk_get_lowest_dev(struct net_device *dev, 2981 struct sock *sk); 2982 struct net_device *dev_get_by_index(struct net *net, int ifindex); 2983 struct net_device *__dev_get_by_index(struct net *net, int ifindex); 2984 struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex); 2985 struct net_device *dev_get_by_napi_id(unsigned int napi_id); 2986 int netdev_get_name(struct net *net, char *name, int ifindex); 2987 int dev_restart(struct net_device *dev); 2988 int skb_gro_receive(struct sk_buff *p, struct sk_buff *skb); 2989 int skb_gro_receive_list(struct sk_buff *p, struct sk_buff *skb); 2990 2991 static inline unsigned int skb_gro_offset(const struct sk_buff *skb) 2992 { 2993 return NAPI_GRO_CB(skb)->data_offset; 2994 } 2995 2996 static inline unsigned int skb_gro_len(const struct sk_buff *skb) 2997 { 2998 return skb->len - NAPI_GRO_CB(skb)->data_offset; 2999 } 3000 3001 static inline void skb_gro_pull(struct sk_buff *skb, unsigned int len) 3002 { 3003 NAPI_GRO_CB(skb)->data_offset += len; 3004 } 3005 3006 static inline void *skb_gro_header_fast(struct sk_buff *skb, 3007 unsigned int offset) 3008 { 3009 return NAPI_GRO_CB(skb)->frag0 + offset; 3010 } 3011 3012 static inline int skb_gro_header_hard(struct sk_buff *skb, unsigned int hlen) 3013 { 3014 return NAPI_GRO_CB(skb)->frag0_len < hlen; 3015 } 3016 3017 static inline void skb_gro_frag0_invalidate(struct sk_buff *skb) 3018 { 3019 NAPI_GRO_CB(skb)->frag0 = NULL; 3020 NAPI_GRO_CB(skb)->frag0_len = 0; 3021 } 3022 3023 static inline void *skb_gro_header_slow(struct sk_buff *skb, unsigned int hlen, 3024 unsigned int offset) 3025 { 3026 if (!pskb_may_pull(skb, hlen)) 3027 return NULL; 3028 3029 skb_gro_frag0_invalidate(skb); 3030 return skb->data + offset; 3031 } 3032 3033 static inline void *skb_gro_network_header(struct sk_buff *skb) 3034 { 3035 return (NAPI_GRO_CB(skb)->frag0 ?: skb->data) + 3036 skb_network_offset(skb); 3037 } 3038 3039 static inline void skb_gro_postpull_rcsum(struct sk_buff *skb, 3040 const void *start, unsigned int len) 3041 { 3042 if (NAPI_GRO_CB(skb)->csum_valid) 3043 NAPI_GRO_CB(skb)->csum = csum_sub(NAPI_GRO_CB(skb)->csum, 3044 csum_partial(start, len, 0)); 3045 } 3046 3047 /* GRO checksum functions. These are logical equivalents of the normal 3048 * checksum functions (in skbuff.h) except that they operate on the GRO 3049 * offsets and fields in sk_buff. 3050 */ 3051 3052 __sum16 __skb_gro_checksum_complete(struct sk_buff *skb); 3053 3054 static inline bool skb_at_gro_remcsum_start(struct sk_buff *skb) 3055 { 3056 return (NAPI_GRO_CB(skb)->gro_remcsum_start == skb_gro_offset(skb)); 3057 } 3058 3059 static inline bool __skb_gro_checksum_validate_needed(struct sk_buff *skb, 3060 bool zero_okay, 3061 __sum16 check) 3062 { 3063 return ((skb->ip_summed != CHECKSUM_PARTIAL || 3064 skb_checksum_start_offset(skb) < 3065 skb_gro_offset(skb)) && 3066 !skb_at_gro_remcsum_start(skb) && 3067 NAPI_GRO_CB(skb)->csum_cnt == 0 && 3068 (!zero_okay || check)); 3069 } 3070 3071 static inline __sum16 __skb_gro_checksum_validate_complete(struct sk_buff *skb, 3072 __wsum psum) 3073 { 3074 if (NAPI_GRO_CB(skb)->csum_valid && 3075 !csum_fold(csum_add(psum, NAPI_GRO_CB(skb)->csum))) 3076 return 0; 3077 3078 NAPI_GRO_CB(skb)->csum = psum; 3079 3080 return __skb_gro_checksum_complete(skb); 3081 } 3082 3083 static inline void skb_gro_incr_csum_unnecessary(struct sk_buff *skb) 3084 { 3085 if (NAPI_GRO_CB(skb)->csum_cnt > 0) { 3086 /* Consume a checksum from CHECKSUM_UNNECESSARY */ 3087 NAPI_GRO_CB(skb)->csum_cnt--; 3088 } else { 3089 /* Update skb for CHECKSUM_UNNECESSARY and csum_level when we 3090 * verified a new top level checksum or an encapsulated one 3091 * during GRO. This saves work if we fallback to normal path. 3092 */ 3093 __skb_incr_checksum_unnecessary(skb); 3094 } 3095 } 3096 3097 #define __skb_gro_checksum_validate(skb, proto, zero_okay, check, \ 3098 compute_pseudo) \ 3099 ({ \ 3100 __sum16 __ret = 0; \ 3101 if (__skb_gro_checksum_validate_needed(skb, zero_okay, check)) \ 3102 __ret = __skb_gro_checksum_validate_complete(skb, \ 3103 compute_pseudo(skb, proto)); \ 3104 if (!__ret) \ 3105 skb_gro_incr_csum_unnecessary(skb); \ 3106 __ret; \ 3107 }) 3108 3109 #define skb_gro_checksum_validate(skb, proto, compute_pseudo) \ 3110 __skb_gro_checksum_validate(skb, proto, false, 0, compute_pseudo) 3111 3112 #define skb_gro_checksum_validate_zero_check(skb, proto, check, \ 3113 compute_pseudo) \ 3114 __skb_gro_checksum_validate(skb, proto, true, check, compute_pseudo) 3115 3116 #define skb_gro_checksum_simple_validate(skb) \ 3117 __skb_gro_checksum_validate(skb, 0, false, 0, null_compute_pseudo) 3118 3119 static inline bool __skb_gro_checksum_convert_check(struct sk_buff *skb) 3120 { 3121 return (NAPI_GRO_CB(skb)->csum_cnt == 0 && 3122 !NAPI_GRO_CB(skb)->csum_valid); 3123 } 3124 3125 static inline void __skb_gro_checksum_convert(struct sk_buff *skb, 3126 __wsum pseudo) 3127 { 3128 NAPI_GRO_CB(skb)->csum = ~pseudo; 3129 NAPI_GRO_CB(skb)->csum_valid = 1; 3130 } 3131 3132 #define skb_gro_checksum_try_convert(skb, proto, compute_pseudo) \ 3133 do { \ 3134 if (__skb_gro_checksum_convert_check(skb)) \ 3135 __skb_gro_checksum_convert(skb, \ 3136 compute_pseudo(skb, proto)); \ 3137 } while (0) 3138 3139 struct gro_remcsum { 3140 int offset; 3141 __wsum delta; 3142 }; 3143 3144 static inline void skb_gro_remcsum_init(struct gro_remcsum *grc) 3145 { 3146 grc->offset = 0; 3147 grc->delta = 0; 3148 } 3149 3150 static inline void *skb_gro_remcsum_process(struct sk_buff *skb, void *ptr, 3151 unsigned int off, size_t hdrlen, 3152 int start, int offset, 3153 struct gro_remcsum *grc, 3154 bool nopartial) 3155 { 3156 __wsum delta; 3157 size_t plen = hdrlen + max_t(size_t, offset + sizeof(u16), start); 3158 3159 BUG_ON(!NAPI_GRO_CB(skb)->csum_valid); 3160 3161 if (!nopartial) { 3162 NAPI_GRO_CB(skb)->gro_remcsum_start = off + hdrlen + start; 3163 return ptr; 3164 } 3165 3166 ptr = skb_gro_header_fast(skb, off); 3167 if (skb_gro_header_hard(skb, off + plen)) { 3168 ptr = skb_gro_header_slow(skb, off + plen, off); 3169 if (!ptr) 3170 return NULL; 3171 } 3172 3173 delta = remcsum_adjust(ptr + hdrlen, NAPI_GRO_CB(skb)->csum, 3174 start, offset); 3175 3176 /* Adjust skb->csum since we changed the packet */ 3177 NAPI_GRO_CB(skb)->csum = csum_add(NAPI_GRO_CB(skb)->csum, delta); 3178 3179 grc->offset = off + hdrlen + offset; 3180 grc->delta = delta; 3181 3182 return ptr; 3183 } 3184 3185 static inline void skb_gro_remcsum_cleanup(struct sk_buff *skb, 3186 struct gro_remcsum *grc) 3187 { 3188 void *ptr; 3189 size_t plen = grc->offset + sizeof(u16); 3190 3191 if (!grc->delta) 3192 return; 3193 3194 ptr = skb_gro_header_fast(skb, grc->offset); 3195 if (skb_gro_header_hard(skb, grc->offset + sizeof(u16))) { 3196 ptr = skb_gro_header_slow(skb, plen, grc->offset); 3197 if (!ptr) 3198 return; 3199 } 3200 3201 remcsum_unadjust((__sum16 *)ptr, grc->delta); 3202 } 3203 3204 #ifdef CONFIG_XFRM_OFFLOAD 3205 static inline void skb_gro_flush_final(struct sk_buff *skb, struct sk_buff *pp, int flush) 3206 { 3207 if (PTR_ERR(pp) != -EINPROGRESS) 3208 NAPI_GRO_CB(skb)->flush |= flush; 3209 } 3210 static inline void skb_gro_flush_final_remcsum(struct sk_buff *skb, 3211 struct sk_buff *pp, 3212 int flush, 3213 struct gro_remcsum *grc) 3214 { 3215 if (PTR_ERR(pp) != -EINPROGRESS) { 3216 NAPI_GRO_CB(skb)->flush |= flush; 3217 skb_gro_remcsum_cleanup(skb, grc); 3218 skb->remcsum_offload = 0; 3219 } 3220 } 3221 #else 3222 static inline void skb_gro_flush_final(struct sk_buff *skb, struct sk_buff *pp, int flush) 3223 { 3224 NAPI_GRO_CB(skb)->flush |= flush; 3225 } 3226 static inline void skb_gro_flush_final_remcsum(struct sk_buff *skb, 3227 struct sk_buff *pp, 3228 int flush, 3229 struct gro_remcsum *grc) 3230 { 3231 NAPI_GRO_CB(skb)->flush |= flush; 3232 skb_gro_remcsum_cleanup(skb, grc); 3233 skb->remcsum_offload = 0; 3234 } 3235 #endif 3236 3237 static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev, 3238 unsigned short type, 3239 const void *daddr, const void *saddr, 3240 unsigned int len) 3241 { 3242 if (!dev->header_ops || !dev->header_ops->create) 3243 return 0; 3244 3245 return dev->header_ops->create(skb, dev, type, daddr, saddr, len); 3246 } 3247 3248 static inline int dev_parse_header(const struct sk_buff *skb, 3249 unsigned char *haddr) 3250 { 3251 const struct net_device *dev = skb->dev; 3252 3253 if (!dev->header_ops || !dev->header_ops->parse) 3254 return 0; 3255 return dev->header_ops->parse(skb, haddr); 3256 } 3257 3258 static inline __be16 dev_parse_header_protocol(const struct sk_buff *skb) 3259 { 3260 const struct net_device *dev = skb->dev; 3261 3262 if (!dev->header_ops || !dev->header_ops->parse_protocol) 3263 return 0; 3264 return dev->header_ops->parse_protocol(skb); 3265 } 3266 3267 /* ll_header must have at least hard_header_len allocated */ 3268 static inline bool dev_validate_header(const struct net_device *dev, 3269 char *ll_header, int len) 3270 { 3271 if (likely(len >= dev->hard_header_len)) 3272 return true; 3273 if (len < dev->min_header_len) 3274 return false; 3275 3276 if (capable(CAP_SYS_RAWIO)) { 3277 memset(ll_header + len, 0, dev->hard_header_len - len); 3278 return true; 3279 } 3280 3281 if (dev->header_ops && dev->header_ops->validate) 3282 return dev->header_ops->validate(ll_header, len); 3283 3284 return false; 3285 } 3286 3287 static inline bool dev_has_header(const struct net_device *dev) 3288 { 3289 return dev->header_ops && dev->header_ops->create; 3290 } 3291 3292 #ifdef CONFIG_NET_FLOW_LIMIT 3293 #define FLOW_LIMIT_HISTORY (1 << 7) /* must be ^2 and !overflow buckets */ 3294 struct sd_flow_limit { 3295 u64 count; 3296 unsigned int num_buckets; 3297 unsigned int history_head; 3298 u16 history[FLOW_LIMIT_HISTORY]; 3299 u8 buckets[]; 3300 }; 3301 3302 extern int netdev_flow_limit_table_len; 3303 #endif /* CONFIG_NET_FLOW_LIMIT */ 3304 3305 /* 3306 * Incoming packets are placed on per-CPU queues 3307 */ 3308 struct softnet_data { 3309 struct list_head poll_list; 3310 struct sk_buff_head process_queue; 3311 3312 /* stats */ 3313 unsigned int processed; 3314 unsigned int time_squeeze; 3315 unsigned int received_rps; 3316 #ifdef CONFIG_RPS 3317 struct softnet_data *rps_ipi_list; 3318 #endif 3319 #ifdef CONFIG_NET_FLOW_LIMIT 3320 struct sd_flow_limit __rcu *flow_limit; 3321 #endif 3322 struct Qdisc *output_queue; 3323 struct Qdisc **output_queue_tailp; 3324 struct sk_buff *completion_queue; 3325 #ifdef CONFIG_XFRM_OFFLOAD 3326 struct sk_buff_head xfrm_backlog; 3327 #endif 3328 /* written and read only by owning cpu: */ 3329 struct { 3330 u16 recursion; 3331 u8 more; 3332 } xmit; 3333 #ifdef CONFIG_RPS 3334 /* input_queue_head should be written by cpu owning this struct, 3335 * and only read by other cpus. Worth using a cache line. 3336 */ 3337 unsigned int input_queue_head ____cacheline_aligned_in_smp; 3338 3339 /* Elements below can be accessed between CPUs for RPS/RFS */ 3340 call_single_data_t csd ____cacheline_aligned_in_smp; 3341 struct softnet_data *rps_ipi_next; 3342 unsigned int cpu; 3343 unsigned int input_queue_tail; 3344 #endif 3345 unsigned int dropped; 3346 struct sk_buff_head input_pkt_queue; 3347 struct napi_struct backlog; 3348 3349 }; 3350 3351 static inline void input_queue_head_incr(struct softnet_data *sd) 3352 { 3353 #ifdef CONFIG_RPS 3354 sd->input_queue_head++; 3355 #endif 3356 } 3357 3358 static inline void input_queue_tail_incr_save(struct softnet_data *sd, 3359 unsigned int *qtail) 3360 { 3361 #ifdef CONFIG_RPS 3362 *qtail = ++sd->input_queue_tail; 3363 #endif 3364 } 3365 3366 DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data); 3367 3368 static inline int dev_recursion_level(void) 3369 { 3370 return this_cpu_read(softnet_data.xmit.recursion); 3371 } 3372 3373 #define XMIT_RECURSION_LIMIT 8 3374 static inline bool dev_xmit_recursion(void) 3375 { 3376 return unlikely(__this_cpu_read(softnet_data.xmit.recursion) > 3377 XMIT_RECURSION_LIMIT); 3378 } 3379 3380 static inline void dev_xmit_recursion_inc(void) 3381 { 3382 __this_cpu_inc(softnet_data.xmit.recursion); 3383 } 3384 3385 static inline void dev_xmit_recursion_dec(void) 3386 { 3387 __this_cpu_dec(softnet_data.xmit.recursion); 3388 } 3389 3390 void __netif_schedule(struct Qdisc *q); 3391 void netif_schedule_queue(struct netdev_queue *txq); 3392 3393 static inline void netif_tx_schedule_all(struct net_device *dev) 3394 { 3395 unsigned int i; 3396 3397 for (i = 0; i < dev->num_tx_queues; i++) 3398 netif_schedule_queue(netdev_get_tx_queue(dev, i)); 3399 } 3400 3401 static __always_inline void netif_tx_start_queue(struct netdev_queue *dev_queue) 3402 { 3403 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state); 3404 } 3405 3406 /** 3407 * netif_start_queue - allow transmit 3408 * @dev: network device 3409 * 3410 * Allow upper layers to call the device hard_start_xmit routine. 3411 */ 3412 static inline void netif_start_queue(struct net_device *dev) 3413 { 3414 netif_tx_start_queue(netdev_get_tx_queue(dev, 0)); 3415 } 3416 3417 static inline void netif_tx_start_all_queues(struct net_device *dev) 3418 { 3419 unsigned int i; 3420 3421 for (i = 0; i < dev->num_tx_queues; i++) { 3422 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 3423 netif_tx_start_queue(txq); 3424 } 3425 } 3426 3427 void netif_tx_wake_queue(struct netdev_queue *dev_queue); 3428 3429 /** 3430 * netif_wake_queue - restart transmit 3431 * @dev: network device 3432 * 3433 * Allow upper layers to call the device hard_start_xmit routine. 3434 * Used for flow control when transmit resources are available. 3435 */ 3436 static inline void netif_wake_queue(struct net_device *dev) 3437 { 3438 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0)); 3439 } 3440 3441 static inline void netif_tx_wake_all_queues(struct net_device *dev) 3442 { 3443 unsigned int i; 3444 3445 for (i = 0; i < dev->num_tx_queues; i++) { 3446 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 3447 netif_tx_wake_queue(txq); 3448 } 3449 } 3450 3451 static __always_inline void netif_tx_stop_queue(struct netdev_queue *dev_queue) 3452 { 3453 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state); 3454 } 3455 3456 /** 3457 * netif_stop_queue - stop transmitted packets 3458 * @dev: network device 3459 * 3460 * Stop upper layers calling the device hard_start_xmit routine. 3461 * Used for flow control when transmit resources are unavailable. 3462 */ 3463 static inline void netif_stop_queue(struct net_device *dev) 3464 { 3465 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0)); 3466 } 3467 3468 void netif_tx_stop_all_queues(struct net_device *dev); 3469 3470 static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue) 3471 { 3472 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state); 3473 } 3474 3475 /** 3476 * netif_queue_stopped - test if transmit queue is flowblocked 3477 * @dev: network device 3478 * 3479 * Test if transmit queue on device is currently unable to send. 3480 */ 3481 static inline bool netif_queue_stopped(const struct net_device *dev) 3482 { 3483 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0)); 3484 } 3485 3486 static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue) 3487 { 3488 return dev_queue->state & QUEUE_STATE_ANY_XOFF; 3489 } 3490 3491 static inline bool 3492 netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue) 3493 { 3494 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN; 3495 } 3496 3497 static inline bool 3498 netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue) 3499 { 3500 return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN; 3501 } 3502 3503 /** 3504 * netdev_queue_set_dql_min_limit - set dql minimum limit 3505 * @dev_queue: pointer to transmit queue 3506 * @min_limit: dql minimum limit 3507 * 3508 * Forces xmit_more() to return true until the minimum threshold 3509 * defined by @min_limit is reached (or until the tx queue is 3510 * empty). Warning: to be use with care, misuse will impact the 3511 * latency. 3512 */ 3513 static inline void netdev_queue_set_dql_min_limit(struct netdev_queue *dev_queue, 3514 unsigned int min_limit) 3515 { 3516 #ifdef CONFIG_BQL 3517 dev_queue->dql.min_limit = min_limit; 3518 #endif 3519 } 3520 3521 /** 3522 * netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write 3523 * @dev_queue: pointer to transmit queue 3524 * 3525 * BQL enabled drivers might use this helper in their ndo_start_xmit(), 3526 * to give appropriate hint to the CPU. 3527 */ 3528 static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue) 3529 { 3530 #ifdef CONFIG_BQL 3531 prefetchw(&dev_queue->dql.num_queued); 3532 #endif 3533 } 3534 3535 /** 3536 * netdev_txq_bql_complete_prefetchw - prefetch bql data for write 3537 * @dev_queue: pointer to transmit queue 3538 * 3539 * BQL enabled drivers might use this helper in their TX completion path, 3540 * to give appropriate hint to the CPU. 3541 */ 3542 static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue) 3543 { 3544 #ifdef CONFIG_BQL 3545 prefetchw(&dev_queue->dql.limit); 3546 #endif 3547 } 3548 3549 static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue, 3550 unsigned int bytes) 3551 { 3552 #ifdef CONFIG_BQL 3553 dql_queued(&dev_queue->dql, bytes); 3554 3555 if (likely(dql_avail(&dev_queue->dql) >= 0)) 3556 return; 3557 3558 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state); 3559 3560 /* 3561 * The XOFF flag must be set before checking the dql_avail below, 3562 * because in netdev_tx_completed_queue we update the dql_completed 3563 * before checking the XOFF flag. 3564 */ 3565 smp_mb(); 3566 3567 /* check again in case another CPU has just made room avail */ 3568 if (unlikely(dql_avail(&dev_queue->dql) >= 0)) 3569 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state); 3570 #endif 3571 } 3572 3573 /* Variant of netdev_tx_sent_queue() for drivers that are aware 3574 * that they should not test BQL status themselves. 3575 * We do want to change __QUEUE_STATE_STACK_XOFF only for the last 3576 * skb of a batch. 3577 * Returns true if the doorbell must be used to kick the NIC. 3578 */ 3579 static inline bool __netdev_tx_sent_queue(struct netdev_queue *dev_queue, 3580 unsigned int bytes, 3581 bool xmit_more) 3582 { 3583 if (xmit_more) { 3584 #ifdef CONFIG_BQL 3585 dql_queued(&dev_queue->dql, bytes); 3586 #endif 3587 return netif_tx_queue_stopped(dev_queue); 3588 } 3589 netdev_tx_sent_queue(dev_queue, bytes); 3590 return true; 3591 } 3592 3593 /** 3594 * netdev_sent_queue - report the number of bytes queued to hardware 3595 * @dev: network device 3596 * @bytes: number of bytes queued to the hardware device queue 3597 * 3598 * Report the number of bytes queued for sending/completion to the network 3599 * device hardware queue. @bytes should be a good approximation and should 3600 * exactly match netdev_completed_queue() @bytes 3601 */ 3602 static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes) 3603 { 3604 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes); 3605 } 3606 3607 static inline bool __netdev_sent_queue(struct net_device *dev, 3608 unsigned int bytes, 3609 bool xmit_more) 3610 { 3611 return __netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes, 3612 xmit_more); 3613 } 3614 3615 static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue, 3616 unsigned int pkts, unsigned int bytes) 3617 { 3618 #ifdef CONFIG_BQL 3619 if (unlikely(!bytes)) 3620 return; 3621 3622 dql_completed(&dev_queue->dql, bytes); 3623 3624 /* 3625 * Without the memory barrier there is a small possiblity that 3626 * netdev_tx_sent_queue will miss the update and cause the queue to 3627 * be stopped forever 3628 */ 3629 smp_mb(); 3630 3631 if (unlikely(dql_avail(&dev_queue->dql) < 0)) 3632 return; 3633 3634 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state)) 3635 netif_schedule_queue(dev_queue); 3636 #endif 3637 } 3638 3639 /** 3640 * netdev_completed_queue - report bytes and packets completed by device 3641 * @dev: network device 3642 * @pkts: actual number of packets sent over the medium 3643 * @bytes: actual number of bytes sent over the medium 3644 * 3645 * Report the number of bytes and packets transmitted by the network device 3646 * hardware queue over the physical medium, @bytes must exactly match the 3647 * @bytes amount passed to netdev_sent_queue() 3648 */ 3649 static inline void netdev_completed_queue(struct net_device *dev, 3650 unsigned int pkts, unsigned int bytes) 3651 { 3652 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes); 3653 } 3654 3655 static inline void netdev_tx_reset_queue(struct netdev_queue *q) 3656 { 3657 #ifdef CONFIG_BQL 3658 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state); 3659 dql_reset(&q->dql); 3660 #endif 3661 } 3662 3663 /** 3664 * netdev_reset_queue - reset the packets and bytes count of a network device 3665 * @dev_queue: network device 3666 * 3667 * Reset the bytes and packet count of a network device and clear the 3668 * software flow control OFF bit for this network device 3669 */ 3670 static inline void netdev_reset_queue(struct net_device *dev_queue) 3671 { 3672 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0)); 3673 } 3674 3675 /** 3676 * netdev_cap_txqueue - check if selected tx queue exceeds device queues 3677 * @dev: network device 3678 * @queue_index: given tx queue index 3679 * 3680 * Returns 0 if given tx queue index >= number of device tx queues, 3681 * otherwise returns the originally passed tx queue index. 3682 */ 3683 static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index) 3684 { 3685 if (unlikely(queue_index >= dev->real_num_tx_queues)) { 3686 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n", 3687 dev->name, queue_index, 3688 dev->real_num_tx_queues); 3689 return 0; 3690 } 3691 3692 return queue_index; 3693 } 3694 3695 /** 3696 * netif_running - test if up 3697 * @dev: network device 3698 * 3699 * Test if the device has been brought up. 3700 */ 3701 static inline bool netif_running(const struct net_device *dev) 3702 { 3703 return test_bit(__LINK_STATE_START, &dev->state); 3704 } 3705 3706 /* 3707 * Routines to manage the subqueues on a device. We only need start, 3708 * stop, and a check if it's stopped. All other device management is 3709 * done at the overall netdevice level. 3710 * Also test the device if we're multiqueue. 3711 */ 3712 3713 /** 3714 * netif_start_subqueue - allow sending packets on subqueue 3715 * @dev: network device 3716 * @queue_index: sub queue index 3717 * 3718 * Start individual transmit queue of a device with multiple transmit queues. 3719 */ 3720 static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index) 3721 { 3722 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 3723 3724 netif_tx_start_queue(txq); 3725 } 3726 3727 /** 3728 * netif_stop_subqueue - stop sending packets on subqueue 3729 * @dev: network device 3730 * @queue_index: sub queue index 3731 * 3732 * Stop individual transmit queue of a device with multiple transmit queues. 3733 */ 3734 static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index) 3735 { 3736 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 3737 netif_tx_stop_queue(txq); 3738 } 3739 3740 /** 3741 * __netif_subqueue_stopped - test status of subqueue 3742 * @dev: network device 3743 * @queue_index: sub queue index 3744 * 3745 * Check individual transmit queue of a device with multiple transmit queues. 3746 */ 3747 static inline bool __netif_subqueue_stopped(const struct net_device *dev, 3748 u16 queue_index) 3749 { 3750 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 3751 3752 return netif_tx_queue_stopped(txq); 3753 } 3754 3755 /** 3756 * netif_subqueue_stopped - test status of subqueue 3757 * @dev: network device 3758 * @skb: sub queue buffer pointer 3759 * 3760 * Check individual transmit queue of a device with multiple transmit queues. 3761 */ 3762 static inline bool netif_subqueue_stopped(const struct net_device *dev, 3763 struct sk_buff *skb) 3764 { 3765 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb)); 3766 } 3767 3768 /** 3769 * netif_wake_subqueue - allow sending packets on subqueue 3770 * @dev: network device 3771 * @queue_index: sub queue index 3772 * 3773 * Resume individual transmit queue of a device with multiple transmit queues. 3774 */ 3775 static inline void netif_wake_subqueue(struct net_device *dev, u16 queue_index) 3776 { 3777 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 3778 3779 netif_tx_wake_queue(txq); 3780 } 3781 3782 #ifdef CONFIG_XPS 3783 int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask, 3784 u16 index); 3785 int __netif_set_xps_queue(struct net_device *dev, const unsigned long *mask, 3786 u16 index, enum xps_map_type type); 3787 3788 /** 3789 * netif_attr_test_mask - Test a CPU or Rx queue set in a mask 3790 * @j: CPU/Rx queue index 3791 * @mask: bitmask of all cpus/rx queues 3792 * @nr_bits: number of bits in the bitmask 3793 * 3794 * Test if a CPU or Rx queue index is set in a mask of all CPU/Rx queues. 3795 */ 3796 static inline bool netif_attr_test_mask(unsigned long j, 3797 const unsigned long *mask, 3798 unsigned int nr_bits) 3799 { 3800 cpu_max_bits_warn(j, nr_bits); 3801 return test_bit(j, mask); 3802 } 3803 3804 /** 3805 * netif_attr_test_online - Test for online CPU/Rx queue 3806 * @j: CPU/Rx queue index 3807 * @online_mask: bitmask for CPUs/Rx queues that are online 3808 * @nr_bits: number of bits in the bitmask 3809 * 3810 * Returns true if a CPU/Rx queue is online. 3811 */ 3812 static inline bool netif_attr_test_online(unsigned long j, 3813 const unsigned long *online_mask, 3814 unsigned int nr_bits) 3815 { 3816 cpu_max_bits_warn(j, nr_bits); 3817 3818 if (online_mask) 3819 return test_bit(j, online_mask); 3820 3821 return (j < nr_bits); 3822 } 3823 3824 /** 3825 * netif_attrmask_next - get the next CPU/Rx queue in a cpu/Rx queues mask 3826 * @n: CPU/Rx queue index 3827 * @srcp: the cpumask/Rx queue mask pointer 3828 * @nr_bits: number of bits in the bitmask 3829 * 3830 * Returns >= nr_bits if no further CPUs/Rx queues set. 3831 */ 3832 static inline unsigned int netif_attrmask_next(int n, const unsigned long *srcp, 3833 unsigned int nr_bits) 3834 { 3835 /* -1 is a legal arg here. */ 3836 if (n != -1) 3837 cpu_max_bits_warn(n, nr_bits); 3838 3839 if (srcp) 3840 return find_next_bit(srcp, nr_bits, n + 1); 3841 3842 return n + 1; 3843 } 3844 3845 /** 3846 * netif_attrmask_next_and - get the next CPU/Rx queue in \*src1p & \*src2p 3847 * @n: CPU/Rx queue index 3848 * @src1p: the first CPUs/Rx queues mask pointer 3849 * @src2p: the second CPUs/Rx queues mask pointer 3850 * @nr_bits: number of bits in the bitmask 3851 * 3852 * Returns >= nr_bits if no further CPUs/Rx queues set in both. 3853 */ 3854 static inline int netif_attrmask_next_and(int n, const unsigned long *src1p, 3855 const unsigned long *src2p, 3856 unsigned int nr_bits) 3857 { 3858 /* -1 is a legal arg here. */ 3859 if (n != -1) 3860 cpu_max_bits_warn(n, nr_bits); 3861 3862 if (src1p && src2p) 3863 return find_next_and_bit(src1p, src2p, nr_bits, n + 1); 3864 else if (src1p) 3865 return find_next_bit(src1p, nr_bits, n + 1); 3866 else if (src2p) 3867 return find_next_bit(src2p, nr_bits, n + 1); 3868 3869 return n + 1; 3870 } 3871 #else 3872 static inline int netif_set_xps_queue(struct net_device *dev, 3873 const struct cpumask *mask, 3874 u16 index) 3875 { 3876 return 0; 3877 } 3878 3879 static inline int __netif_set_xps_queue(struct net_device *dev, 3880 const unsigned long *mask, 3881 u16 index, enum xps_map_type type) 3882 { 3883 return 0; 3884 } 3885 #endif 3886 3887 /** 3888 * netif_is_multiqueue - test if device has multiple transmit queues 3889 * @dev: network device 3890 * 3891 * Check if device has multiple transmit queues 3892 */ 3893 static inline bool netif_is_multiqueue(const struct net_device *dev) 3894 { 3895 return dev->num_tx_queues > 1; 3896 } 3897 3898 int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq); 3899 3900 #ifdef CONFIG_SYSFS 3901 int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq); 3902 #else 3903 static inline int netif_set_real_num_rx_queues(struct net_device *dev, 3904 unsigned int rxqs) 3905 { 3906 dev->real_num_rx_queues = rxqs; 3907 return 0; 3908 } 3909 #endif 3910 3911 static inline struct netdev_rx_queue * 3912 __netif_get_rx_queue(struct net_device *dev, unsigned int rxq) 3913 { 3914 return dev->_rx + rxq; 3915 } 3916 3917 #ifdef CONFIG_SYSFS 3918 static inline unsigned int get_netdev_rx_queue_index( 3919 struct netdev_rx_queue *queue) 3920 { 3921 struct net_device *dev = queue->dev; 3922 int index = queue - dev->_rx; 3923 3924 BUG_ON(index >= dev->num_rx_queues); 3925 return index; 3926 } 3927 #endif 3928 3929 #define DEFAULT_MAX_NUM_RSS_QUEUES (8) 3930 int netif_get_num_default_rss_queues(void); 3931 3932 enum skb_free_reason { 3933 SKB_REASON_CONSUMED, 3934 SKB_REASON_DROPPED, 3935 }; 3936 3937 void __dev_kfree_skb_irq(struct sk_buff *skb, enum skb_free_reason reason); 3938 void __dev_kfree_skb_any(struct sk_buff *skb, enum skb_free_reason reason); 3939 3940 /* 3941 * It is not allowed to call kfree_skb() or consume_skb() from hardware 3942 * interrupt context or with hardware interrupts being disabled. 3943 * (in_irq() || irqs_disabled()) 3944 * 3945 * We provide four helpers that can be used in following contexts : 3946 * 3947 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context, 3948 * replacing kfree_skb(skb) 3949 * 3950 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context. 3951 * Typically used in place of consume_skb(skb) in TX completion path 3952 * 3953 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context, 3954 * replacing kfree_skb(skb) 3955 * 3956 * dev_consume_skb_any(skb) when caller doesn't know its current irq context, 3957 * and consumed a packet. Used in place of consume_skb(skb) 3958 */ 3959 static inline void dev_kfree_skb_irq(struct sk_buff *skb) 3960 { 3961 __dev_kfree_skb_irq(skb, SKB_REASON_DROPPED); 3962 } 3963 3964 static inline void dev_consume_skb_irq(struct sk_buff *skb) 3965 { 3966 __dev_kfree_skb_irq(skb, SKB_REASON_CONSUMED); 3967 } 3968 3969 static inline void dev_kfree_skb_any(struct sk_buff *skb) 3970 { 3971 __dev_kfree_skb_any(skb, SKB_REASON_DROPPED); 3972 } 3973 3974 static inline void dev_consume_skb_any(struct sk_buff *skb) 3975 { 3976 __dev_kfree_skb_any(skb, SKB_REASON_CONSUMED); 3977 } 3978 3979 u32 bpf_prog_run_generic_xdp(struct sk_buff *skb, struct xdp_buff *xdp, 3980 struct bpf_prog *xdp_prog); 3981 void generic_xdp_tx(struct sk_buff *skb, struct bpf_prog *xdp_prog); 3982 int do_xdp_generic(struct bpf_prog *xdp_prog, struct sk_buff *skb); 3983 int netif_rx(struct sk_buff *skb); 3984 int netif_rx_ni(struct sk_buff *skb); 3985 int netif_rx_any_context(struct sk_buff *skb); 3986 int netif_receive_skb(struct sk_buff *skb); 3987 int netif_receive_skb_core(struct sk_buff *skb); 3988 void netif_receive_skb_list(struct list_head *head); 3989 gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb); 3990 void napi_gro_flush(struct napi_struct *napi, bool flush_old); 3991 struct sk_buff *napi_get_frags(struct napi_struct *napi); 3992 gro_result_t napi_gro_frags(struct napi_struct *napi); 3993 struct packet_offload *gro_find_receive_by_type(__be16 type); 3994 struct packet_offload *gro_find_complete_by_type(__be16 type); 3995 3996 static inline void napi_free_frags(struct napi_struct *napi) 3997 { 3998 kfree_skb(napi->skb); 3999 napi->skb = NULL; 4000 } 4001 4002 bool netdev_is_rx_handler_busy(struct net_device *dev); 4003 int netdev_rx_handler_register(struct net_device *dev, 4004 rx_handler_func_t *rx_handler, 4005 void *rx_handler_data); 4006 void netdev_rx_handler_unregister(struct net_device *dev); 4007 4008 bool dev_valid_name(const char *name); 4009 int get_user_ifreq(struct ifreq *ifr, void __user **ifrdata, void __user *arg); 4010 int put_user_ifreq(struct ifreq *ifr, void __user *arg); 4011 int dev_ioctl(struct net *net, unsigned int cmd, struct ifreq *ifr, 4012 bool *need_copyout); 4013 int dev_ifconf(struct net *net, struct ifconf __user *ifc); 4014 int dev_ethtool(struct net *net, struct ifreq *); 4015 unsigned int dev_get_flags(const struct net_device *); 4016 int __dev_change_flags(struct net_device *dev, unsigned int flags, 4017 struct netlink_ext_ack *extack); 4018 int dev_change_flags(struct net_device *dev, unsigned int flags, 4019 struct netlink_ext_ack *extack); 4020 void __dev_notify_flags(struct net_device *, unsigned int old_flags, 4021 unsigned int gchanges); 4022 int dev_change_name(struct net_device *, const char *); 4023 int dev_set_alias(struct net_device *, const char *, size_t); 4024 int dev_get_alias(const struct net_device *, char *, size_t); 4025 int __dev_change_net_namespace(struct net_device *dev, struct net *net, 4026 const char *pat, int new_ifindex); 4027 static inline 4028 int dev_change_net_namespace(struct net_device *dev, struct net *net, 4029 const char *pat) 4030 { 4031 return __dev_change_net_namespace(dev, net, pat, 0); 4032 } 4033 int __dev_set_mtu(struct net_device *, int); 4034 int dev_validate_mtu(struct net_device *dev, int mtu, 4035 struct netlink_ext_ack *extack); 4036 int dev_set_mtu_ext(struct net_device *dev, int mtu, 4037 struct netlink_ext_ack *extack); 4038 int dev_set_mtu(struct net_device *, int); 4039 int dev_change_tx_queue_len(struct net_device *, unsigned long); 4040 void dev_set_group(struct net_device *, int); 4041 int dev_pre_changeaddr_notify(struct net_device *dev, const char *addr, 4042 struct netlink_ext_ack *extack); 4043 int dev_set_mac_address(struct net_device *dev, struct sockaddr *sa, 4044 struct netlink_ext_ack *extack); 4045 int dev_set_mac_address_user(struct net_device *dev, struct sockaddr *sa, 4046 struct netlink_ext_ack *extack); 4047 int dev_get_mac_address(struct sockaddr *sa, struct net *net, char *dev_name); 4048 int dev_change_carrier(struct net_device *, bool new_carrier); 4049 int dev_get_phys_port_id(struct net_device *dev, 4050 struct netdev_phys_item_id *ppid); 4051 int dev_get_phys_port_name(struct net_device *dev, 4052 char *name, size_t len); 4053 int dev_get_port_parent_id(struct net_device *dev, 4054 struct netdev_phys_item_id *ppid, bool recurse); 4055 bool netdev_port_same_parent_id(struct net_device *a, struct net_device *b); 4056 int dev_change_proto_down(struct net_device *dev, bool proto_down); 4057 int dev_change_proto_down_generic(struct net_device *dev, bool proto_down); 4058 void dev_change_proto_down_reason(struct net_device *dev, unsigned long mask, 4059 u32 value); 4060 struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev, bool *again); 4061 struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev, 4062 struct netdev_queue *txq, int *ret); 4063 4064 typedef int (*bpf_op_t)(struct net_device *dev, struct netdev_bpf *bpf); 4065 int dev_change_xdp_fd(struct net_device *dev, struct netlink_ext_ack *extack, 4066 int fd, int expected_fd, u32 flags); 4067 int bpf_xdp_link_attach(const union bpf_attr *attr, struct bpf_prog *prog); 4068 u32 dev_xdp_prog_id(struct net_device *dev, enum bpf_xdp_mode mode); 4069 4070 int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb); 4071 int dev_forward_skb(struct net_device *dev, struct sk_buff *skb); 4072 int dev_forward_skb_nomtu(struct net_device *dev, struct sk_buff *skb); 4073 bool is_skb_forwardable(const struct net_device *dev, 4074 const struct sk_buff *skb); 4075 4076 static __always_inline bool __is_skb_forwardable(const struct net_device *dev, 4077 const struct sk_buff *skb, 4078 const bool check_mtu) 4079 { 4080 const u32 vlan_hdr_len = 4; /* VLAN_HLEN */ 4081 unsigned int len; 4082 4083 if (!(dev->flags & IFF_UP)) 4084 return false; 4085 4086 if (!check_mtu) 4087 return true; 4088 4089 len = dev->mtu + dev->hard_header_len + vlan_hdr_len; 4090 if (skb->len <= len) 4091 return true; 4092 4093 /* if TSO is enabled, we don't care about the length as the packet 4094 * could be forwarded without being segmented before 4095 */ 4096 if (skb_is_gso(skb)) 4097 return true; 4098 4099 return false; 4100 } 4101 4102 static __always_inline int ____dev_forward_skb(struct net_device *dev, 4103 struct sk_buff *skb, 4104 const bool check_mtu) 4105 { 4106 if (skb_orphan_frags(skb, GFP_ATOMIC) || 4107 unlikely(!__is_skb_forwardable(dev, skb, check_mtu))) { 4108 atomic_long_inc(&dev->rx_dropped); 4109 kfree_skb(skb); 4110 return NET_RX_DROP; 4111 } 4112 4113 skb_scrub_packet(skb, !net_eq(dev_net(dev), dev_net(skb->dev))); 4114 skb->priority = 0; 4115 return 0; 4116 } 4117 4118 bool dev_nit_active(struct net_device *dev); 4119 void dev_queue_xmit_nit(struct sk_buff *skb, struct net_device *dev); 4120 4121 extern int netdev_budget; 4122 extern unsigned int netdev_budget_usecs; 4123 4124 /* Called by rtnetlink.c:rtnl_unlock() */ 4125 void netdev_run_todo(void); 4126 4127 /** 4128 * dev_put - release reference to device 4129 * @dev: network device 4130 * 4131 * Release reference to device to allow it to be freed. 4132 */ 4133 static inline void dev_put(struct net_device *dev) 4134 { 4135 #ifdef CONFIG_PCPU_DEV_REFCNT 4136 this_cpu_dec(*dev->pcpu_refcnt); 4137 #else 4138 refcount_dec(&dev->dev_refcnt); 4139 #endif 4140 } 4141 4142 /** 4143 * dev_hold - get reference to device 4144 * @dev: network device 4145 * 4146 * Hold reference to device to keep it from being freed. 4147 */ 4148 static inline void dev_hold(struct net_device *dev) 4149 { 4150 #ifdef CONFIG_PCPU_DEV_REFCNT 4151 this_cpu_inc(*dev->pcpu_refcnt); 4152 #else 4153 refcount_inc(&dev->dev_refcnt); 4154 #endif 4155 } 4156 4157 /* Carrier loss detection, dial on demand. The functions netif_carrier_on 4158 * and _off may be called from IRQ context, but it is caller 4159 * who is responsible for serialization of these calls. 4160 * 4161 * The name carrier is inappropriate, these functions should really be 4162 * called netif_lowerlayer_*() because they represent the state of any 4163 * kind of lower layer not just hardware media. 4164 */ 4165 4166 void linkwatch_init_dev(struct net_device *dev); 4167 void linkwatch_fire_event(struct net_device *dev); 4168 void linkwatch_forget_dev(struct net_device *dev); 4169 4170 /** 4171 * netif_carrier_ok - test if carrier present 4172 * @dev: network device 4173 * 4174 * Check if carrier is present on device 4175 */ 4176 static inline bool netif_carrier_ok(const struct net_device *dev) 4177 { 4178 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state); 4179 } 4180 4181 unsigned long dev_trans_start(struct net_device *dev); 4182 4183 void __netdev_watchdog_up(struct net_device *dev); 4184 4185 void netif_carrier_on(struct net_device *dev); 4186 void netif_carrier_off(struct net_device *dev); 4187 void netif_carrier_event(struct net_device *dev); 4188 4189 /** 4190 * netif_dormant_on - mark device as dormant. 4191 * @dev: network device 4192 * 4193 * Mark device as dormant (as per RFC2863). 4194 * 4195 * The dormant state indicates that the relevant interface is not 4196 * actually in a condition to pass packets (i.e., it is not 'up') but is 4197 * in a "pending" state, waiting for some external event. For "on- 4198 * demand" interfaces, this new state identifies the situation where the 4199 * interface is waiting for events to place it in the up state. 4200 */ 4201 static inline void netif_dormant_on(struct net_device *dev) 4202 { 4203 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state)) 4204 linkwatch_fire_event(dev); 4205 } 4206 4207 /** 4208 * netif_dormant_off - set device as not dormant. 4209 * @dev: network device 4210 * 4211 * Device is not in dormant state. 4212 */ 4213 static inline void netif_dormant_off(struct net_device *dev) 4214 { 4215 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state)) 4216 linkwatch_fire_event(dev); 4217 } 4218 4219 /** 4220 * netif_dormant - test if device is dormant 4221 * @dev: network device 4222 * 4223 * Check if device is dormant. 4224 */ 4225 static inline bool netif_dormant(const struct net_device *dev) 4226 { 4227 return test_bit(__LINK_STATE_DORMANT, &dev->state); 4228 } 4229 4230 4231 /** 4232 * netif_testing_on - mark device as under test. 4233 * @dev: network device 4234 * 4235 * Mark device as under test (as per RFC2863). 4236 * 4237 * The testing state indicates that some test(s) must be performed on 4238 * the interface. After completion, of the test, the interface state 4239 * will change to up, dormant, or down, as appropriate. 4240 */ 4241 static inline void netif_testing_on(struct net_device *dev) 4242 { 4243 if (!test_and_set_bit(__LINK_STATE_TESTING, &dev->state)) 4244 linkwatch_fire_event(dev); 4245 } 4246 4247 /** 4248 * netif_testing_off - set device as not under test. 4249 * @dev: network device 4250 * 4251 * Device is not in testing state. 4252 */ 4253 static inline void netif_testing_off(struct net_device *dev) 4254 { 4255 if (test_and_clear_bit(__LINK_STATE_TESTING, &dev->state)) 4256 linkwatch_fire_event(dev); 4257 } 4258 4259 /** 4260 * netif_testing - test if device is under test 4261 * @dev: network device 4262 * 4263 * Check if device is under test 4264 */ 4265 static inline bool netif_testing(const struct net_device *dev) 4266 { 4267 return test_bit(__LINK_STATE_TESTING, &dev->state); 4268 } 4269 4270 4271 /** 4272 * netif_oper_up - test if device is operational 4273 * @dev: network device 4274 * 4275 * Check if carrier is operational 4276 */ 4277 static inline bool netif_oper_up(const struct net_device *dev) 4278 { 4279 return (dev->operstate == IF_OPER_UP || 4280 dev->operstate == IF_OPER_UNKNOWN /* backward compat */); 4281 } 4282 4283 /** 4284 * netif_device_present - is device available or removed 4285 * @dev: network device 4286 * 4287 * Check if device has not been removed from system. 4288 */ 4289 static inline bool netif_device_present(const struct net_device *dev) 4290 { 4291 return test_bit(__LINK_STATE_PRESENT, &dev->state); 4292 } 4293 4294 void netif_device_detach(struct net_device *dev); 4295 4296 void netif_device_attach(struct net_device *dev); 4297 4298 /* 4299 * Network interface message level settings 4300 */ 4301 4302 enum { 4303 NETIF_MSG_DRV_BIT, 4304 NETIF_MSG_PROBE_BIT, 4305 NETIF_MSG_LINK_BIT, 4306 NETIF_MSG_TIMER_BIT, 4307 NETIF_MSG_IFDOWN_BIT, 4308 NETIF_MSG_IFUP_BIT, 4309 NETIF_MSG_RX_ERR_BIT, 4310 NETIF_MSG_TX_ERR_BIT, 4311 NETIF_MSG_TX_QUEUED_BIT, 4312 NETIF_MSG_INTR_BIT, 4313 NETIF_MSG_TX_DONE_BIT, 4314 NETIF_MSG_RX_STATUS_BIT, 4315 NETIF_MSG_PKTDATA_BIT, 4316 NETIF_MSG_HW_BIT, 4317 NETIF_MSG_WOL_BIT, 4318 4319 /* When you add a new bit above, update netif_msg_class_names array 4320 * in net/ethtool/common.c 4321 */ 4322 NETIF_MSG_CLASS_COUNT, 4323 }; 4324 /* Both ethtool_ops interface and internal driver implementation use u32 */ 4325 static_assert(NETIF_MSG_CLASS_COUNT <= 32); 4326 4327 #define __NETIF_MSG_BIT(bit) ((u32)1 << (bit)) 4328 #define __NETIF_MSG(name) __NETIF_MSG_BIT(NETIF_MSG_ ## name ## _BIT) 4329 4330 #define NETIF_MSG_DRV __NETIF_MSG(DRV) 4331 #define NETIF_MSG_PROBE __NETIF_MSG(PROBE) 4332 #define NETIF_MSG_LINK __NETIF_MSG(LINK) 4333 #define NETIF_MSG_TIMER __NETIF_MSG(TIMER) 4334 #define NETIF_MSG_IFDOWN __NETIF_MSG(IFDOWN) 4335 #define NETIF_MSG_IFUP __NETIF_MSG(IFUP) 4336 #define NETIF_MSG_RX_ERR __NETIF_MSG(RX_ERR) 4337 #define NETIF_MSG_TX_ERR __NETIF_MSG(TX_ERR) 4338 #define NETIF_MSG_TX_QUEUED __NETIF_MSG(TX_QUEUED) 4339 #define NETIF_MSG_INTR __NETIF_MSG(INTR) 4340 #define NETIF_MSG_TX_DONE __NETIF_MSG(TX_DONE) 4341 #define NETIF_MSG_RX_STATUS __NETIF_MSG(RX_STATUS) 4342 #define NETIF_MSG_PKTDATA __NETIF_MSG(PKTDATA) 4343 #define NETIF_MSG_HW __NETIF_MSG(HW) 4344 #define NETIF_MSG_WOL __NETIF_MSG(WOL) 4345 4346 #define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV) 4347 #define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE) 4348 #define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK) 4349 #define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER) 4350 #define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN) 4351 #define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP) 4352 #define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR) 4353 #define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR) 4354 #define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED) 4355 #define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR) 4356 #define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE) 4357 #define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS) 4358 #define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA) 4359 #define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW) 4360 #define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL) 4361 4362 static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits) 4363 { 4364 /* use default */ 4365 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8)) 4366 return default_msg_enable_bits; 4367 if (debug_value == 0) /* no output */ 4368 return 0; 4369 /* set low N bits */ 4370 return (1U << debug_value) - 1; 4371 } 4372 4373 static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu) 4374 { 4375 spin_lock(&txq->_xmit_lock); 4376 txq->xmit_lock_owner = cpu; 4377 } 4378 4379 static inline bool __netif_tx_acquire(struct netdev_queue *txq) 4380 { 4381 __acquire(&txq->_xmit_lock); 4382 return true; 4383 } 4384 4385 static inline void __netif_tx_release(struct netdev_queue *txq) 4386 { 4387 __release(&txq->_xmit_lock); 4388 } 4389 4390 static inline void __netif_tx_lock_bh(struct netdev_queue *txq) 4391 { 4392 spin_lock_bh(&txq->_xmit_lock); 4393 txq->xmit_lock_owner = smp_processor_id(); 4394 } 4395 4396 static inline bool __netif_tx_trylock(struct netdev_queue *txq) 4397 { 4398 bool ok = spin_trylock(&txq->_xmit_lock); 4399 if (likely(ok)) 4400 txq->xmit_lock_owner = smp_processor_id(); 4401 return ok; 4402 } 4403 4404 static inline void __netif_tx_unlock(struct netdev_queue *txq) 4405 { 4406 txq->xmit_lock_owner = -1; 4407 spin_unlock(&txq->_xmit_lock); 4408 } 4409 4410 static inline void __netif_tx_unlock_bh(struct netdev_queue *txq) 4411 { 4412 txq->xmit_lock_owner = -1; 4413 spin_unlock_bh(&txq->_xmit_lock); 4414 } 4415 4416 static inline void txq_trans_update(struct netdev_queue *txq) 4417 { 4418 if (txq->xmit_lock_owner != -1) 4419 txq->trans_start = jiffies; 4420 } 4421 4422 /* legacy drivers only, netdev_start_xmit() sets txq->trans_start */ 4423 static inline void netif_trans_update(struct net_device *dev) 4424 { 4425 struct netdev_queue *txq = netdev_get_tx_queue(dev, 0); 4426 4427 if (txq->trans_start != jiffies) 4428 txq->trans_start = jiffies; 4429 } 4430 4431 /** 4432 * netif_tx_lock - grab network device transmit lock 4433 * @dev: network device 4434 * 4435 * Get network device transmit lock 4436 */ 4437 static inline void netif_tx_lock(struct net_device *dev) 4438 { 4439 unsigned int i; 4440 int cpu; 4441 4442 spin_lock(&dev->tx_global_lock); 4443 cpu = smp_processor_id(); 4444 for (i = 0; i < dev->num_tx_queues; i++) { 4445 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 4446 4447 /* We are the only thread of execution doing a 4448 * freeze, but we have to grab the _xmit_lock in 4449 * order to synchronize with threads which are in 4450 * the ->hard_start_xmit() handler and already 4451 * checked the frozen bit. 4452 */ 4453 __netif_tx_lock(txq, cpu); 4454 set_bit(__QUEUE_STATE_FROZEN, &txq->state); 4455 __netif_tx_unlock(txq); 4456 } 4457 } 4458 4459 static inline void netif_tx_lock_bh(struct net_device *dev) 4460 { 4461 local_bh_disable(); 4462 netif_tx_lock(dev); 4463 } 4464 4465 static inline void netif_tx_unlock(struct net_device *dev) 4466 { 4467 unsigned int i; 4468 4469 for (i = 0; i < dev->num_tx_queues; i++) { 4470 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 4471 4472 /* No need to grab the _xmit_lock here. If the 4473 * queue is not stopped for another reason, we 4474 * force a schedule. 4475 */ 4476 clear_bit(__QUEUE_STATE_FROZEN, &txq->state); 4477 netif_schedule_queue(txq); 4478 } 4479 spin_unlock(&dev->tx_global_lock); 4480 } 4481 4482 static inline void netif_tx_unlock_bh(struct net_device *dev) 4483 { 4484 netif_tx_unlock(dev); 4485 local_bh_enable(); 4486 } 4487 4488 #define HARD_TX_LOCK(dev, txq, cpu) { \ 4489 if ((dev->features & NETIF_F_LLTX) == 0) { \ 4490 __netif_tx_lock(txq, cpu); \ 4491 } else { \ 4492 __netif_tx_acquire(txq); \ 4493 } \ 4494 } 4495 4496 #define HARD_TX_TRYLOCK(dev, txq) \ 4497 (((dev->features & NETIF_F_LLTX) == 0) ? \ 4498 __netif_tx_trylock(txq) : \ 4499 __netif_tx_acquire(txq)) 4500 4501 #define HARD_TX_UNLOCK(dev, txq) { \ 4502 if ((dev->features & NETIF_F_LLTX) == 0) { \ 4503 __netif_tx_unlock(txq); \ 4504 } else { \ 4505 __netif_tx_release(txq); \ 4506 } \ 4507 } 4508 4509 static inline void netif_tx_disable(struct net_device *dev) 4510 { 4511 unsigned int i; 4512 int cpu; 4513 4514 local_bh_disable(); 4515 cpu = smp_processor_id(); 4516 spin_lock(&dev->tx_global_lock); 4517 for (i = 0; i < dev->num_tx_queues; i++) { 4518 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 4519 4520 __netif_tx_lock(txq, cpu); 4521 netif_tx_stop_queue(txq); 4522 __netif_tx_unlock(txq); 4523 } 4524 spin_unlock(&dev->tx_global_lock); 4525 local_bh_enable(); 4526 } 4527 4528 static inline void netif_addr_lock(struct net_device *dev) 4529 { 4530 unsigned char nest_level = 0; 4531 4532 #ifdef CONFIG_LOCKDEP 4533 nest_level = dev->nested_level; 4534 #endif 4535 spin_lock_nested(&dev->addr_list_lock, nest_level); 4536 } 4537 4538 static inline void netif_addr_lock_bh(struct net_device *dev) 4539 { 4540 unsigned char nest_level = 0; 4541 4542 #ifdef CONFIG_LOCKDEP 4543 nest_level = dev->nested_level; 4544 #endif 4545 local_bh_disable(); 4546 spin_lock_nested(&dev->addr_list_lock, nest_level); 4547 } 4548 4549 static inline void netif_addr_unlock(struct net_device *dev) 4550 { 4551 spin_unlock(&dev->addr_list_lock); 4552 } 4553 4554 static inline void netif_addr_unlock_bh(struct net_device *dev) 4555 { 4556 spin_unlock_bh(&dev->addr_list_lock); 4557 } 4558 4559 /* 4560 * dev_addrs walker. Should be used only for read access. Call with 4561 * rcu_read_lock held. 4562 */ 4563 #define for_each_dev_addr(dev, ha) \ 4564 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list) 4565 4566 /* These functions live elsewhere (drivers/net/net_init.c, but related) */ 4567 4568 void ether_setup(struct net_device *dev); 4569 4570 /* Support for loadable net-drivers */ 4571 struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name, 4572 unsigned char name_assign_type, 4573 void (*setup)(struct net_device *), 4574 unsigned int txqs, unsigned int rxqs); 4575 #define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \ 4576 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1) 4577 4578 #define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \ 4579 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \ 4580 count) 4581 4582 int register_netdev(struct net_device *dev); 4583 void unregister_netdev(struct net_device *dev); 4584 4585 int devm_register_netdev(struct device *dev, struct net_device *ndev); 4586 4587 /* General hardware address lists handling functions */ 4588 int __hw_addr_sync(struct netdev_hw_addr_list *to_list, 4589 struct netdev_hw_addr_list *from_list, int addr_len); 4590 void __hw_addr_unsync(struct netdev_hw_addr_list *to_list, 4591 struct netdev_hw_addr_list *from_list, int addr_len); 4592 int __hw_addr_sync_dev(struct netdev_hw_addr_list *list, 4593 struct net_device *dev, 4594 int (*sync)(struct net_device *, const unsigned char *), 4595 int (*unsync)(struct net_device *, 4596 const unsigned char *)); 4597 int __hw_addr_ref_sync_dev(struct netdev_hw_addr_list *list, 4598 struct net_device *dev, 4599 int (*sync)(struct net_device *, 4600 const unsigned char *, int), 4601 int (*unsync)(struct net_device *, 4602 const unsigned char *, int)); 4603 void __hw_addr_ref_unsync_dev(struct netdev_hw_addr_list *list, 4604 struct net_device *dev, 4605 int (*unsync)(struct net_device *, 4606 const unsigned char *, int)); 4607 void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list, 4608 struct net_device *dev, 4609 int (*unsync)(struct net_device *, 4610 const unsigned char *)); 4611 void __hw_addr_init(struct netdev_hw_addr_list *list); 4612 4613 /* Functions used for device addresses handling */ 4614 int dev_addr_add(struct net_device *dev, const unsigned char *addr, 4615 unsigned char addr_type); 4616 int dev_addr_del(struct net_device *dev, const unsigned char *addr, 4617 unsigned char addr_type); 4618 void dev_addr_flush(struct net_device *dev); 4619 int dev_addr_init(struct net_device *dev); 4620 4621 /* Functions used for unicast addresses handling */ 4622 int dev_uc_add(struct net_device *dev, const unsigned char *addr); 4623 int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr); 4624 int dev_uc_del(struct net_device *dev, const unsigned char *addr); 4625 int dev_uc_sync(struct net_device *to, struct net_device *from); 4626 int dev_uc_sync_multiple(struct net_device *to, struct net_device *from); 4627 void dev_uc_unsync(struct net_device *to, struct net_device *from); 4628 void dev_uc_flush(struct net_device *dev); 4629 void dev_uc_init(struct net_device *dev); 4630 4631 /** 4632 * __dev_uc_sync - Synchonize device's unicast list 4633 * @dev: device to sync 4634 * @sync: function to call if address should be added 4635 * @unsync: function to call if address should be removed 4636 * 4637 * Add newly added addresses to the interface, and release 4638 * addresses that have been deleted. 4639 */ 4640 static inline int __dev_uc_sync(struct net_device *dev, 4641 int (*sync)(struct net_device *, 4642 const unsigned char *), 4643 int (*unsync)(struct net_device *, 4644 const unsigned char *)) 4645 { 4646 return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync); 4647 } 4648 4649 /** 4650 * __dev_uc_unsync - Remove synchronized addresses from device 4651 * @dev: device to sync 4652 * @unsync: function to call if address should be removed 4653 * 4654 * Remove all addresses that were added to the device by dev_uc_sync(). 4655 */ 4656 static inline void __dev_uc_unsync(struct net_device *dev, 4657 int (*unsync)(struct net_device *, 4658 const unsigned char *)) 4659 { 4660 __hw_addr_unsync_dev(&dev->uc, dev, unsync); 4661 } 4662 4663 /* Functions used for multicast addresses handling */ 4664 int dev_mc_add(struct net_device *dev, const unsigned char *addr); 4665 int dev_mc_add_global(struct net_device *dev, const unsigned char *addr); 4666 int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr); 4667 int dev_mc_del(struct net_device *dev, const unsigned char *addr); 4668 int dev_mc_del_global(struct net_device *dev, const unsigned char *addr); 4669 int dev_mc_sync(struct net_device *to, struct net_device *from); 4670 int dev_mc_sync_multiple(struct net_device *to, struct net_device *from); 4671 void dev_mc_unsync(struct net_device *to, struct net_device *from); 4672 void dev_mc_flush(struct net_device *dev); 4673 void dev_mc_init(struct net_device *dev); 4674 4675 /** 4676 * __dev_mc_sync - Synchonize device's multicast list 4677 * @dev: device to sync 4678 * @sync: function to call if address should be added 4679 * @unsync: function to call if address should be removed 4680 * 4681 * Add newly added addresses to the interface, and release 4682 * addresses that have been deleted. 4683 */ 4684 static inline int __dev_mc_sync(struct net_device *dev, 4685 int (*sync)(struct net_device *, 4686 const unsigned char *), 4687 int (*unsync)(struct net_device *, 4688 const unsigned char *)) 4689 { 4690 return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync); 4691 } 4692 4693 /** 4694 * __dev_mc_unsync - Remove synchronized addresses from device 4695 * @dev: device to sync 4696 * @unsync: function to call if address should be removed 4697 * 4698 * Remove all addresses that were added to the device by dev_mc_sync(). 4699 */ 4700 static inline void __dev_mc_unsync(struct net_device *dev, 4701 int (*unsync)(struct net_device *, 4702 const unsigned char *)) 4703 { 4704 __hw_addr_unsync_dev(&dev->mc, dev, unsync); 4705 } 4706 4707 /* Functions used for secondary unicast and multicast support */ 4708 void dev_set_rx_mode(struct net_device *dev); 4709 void __dev_set_rx_mode(struct net_device *dev); 4710 int dev_set_promiscuity(struct net_device *dev, int inc); 4711 int dev_set_allmulti(struct net_device *dev, int inc); 4712 void netdev_state_change(struct net_device *dev); 4713 void __netdev_notify_peers(struct net_device *dev); 4714 void netdev_notify_peers(struct net_device *dev); 4715 void netdev_features_change(struct net_device *dev); 4716 /* Load a device via the kmod */ 4717 void dev_load(struct net *net, const char *name); 4718 struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev, 4719 struct rtnl_link_stats64 *storage); 4720 void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64, 4721 const struct net_device_stats *netdev_stats); 4722 void dev_fetch_sw_netstats(struct rtnl_link_stats64 *s, 4723 const struct pcpu_sw_netstats __percpu *netstats); 4724 void dev_get_tstats64(struct net_device *dev, struct rtnl_link_stats64 *s); 4725 4726 extern int netdev_max_backlog; 4727 extern int netdev_tstamp_prequeue; 4728 extern int netdev_unregister_timeout_secs; 4729 extern int weight_p; 4730 extern int dev_weight_rx_bias; 4731 extern int dev_weight_tx_bias; 4732 extern int dev_rx_weight; 4733 extern int dev_tx_weight; 4734 extern int gro_normal_batch; 4735 4736 enum { 4737 NESTED_SYNC_IMM_BIT, 4738 NESTED_SYNC_TODO_BIT, 4739 }; 4740 4741 #define __NESTED_SYNC_BIT(bit) ((u32)1 << (bit)) 4742 #define __NESTED_SYNC(name) __NESTED_SYNC_BIT(NESTED_SYNC_ ## name ## _BIT) 4743 4744 #define NESTED_SYNC_IMM __NESTED_SYNC(IMM) 4745 #define NESTED_SYNC_TODO __NESTED_SYNC(TODO) 4746 4747 struct netdev_nested_priv { 4748 unsigned char flags; 4749 void *data; 4750 }; 4751 4752 bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev); 4753 struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev, 4754 struct list_head **iter); 4755 struct net_device *netdev_all_upper_get_next_dev_rcu(struct net_device *dev, 4756 struct list_head **iter); 4757 4758 #ifdef CONFIG_LOCKDEP 4759 static LIST_HEAD(net_unlink_list); 4760 4761 static inline void net_unlink_todo(struct net_device *dev) 4762 { 4763 if (list_empty(&dev->unlink_list)) 4764 list_add_tail(&dev->unlink_list, &net_unlink_list); 4765 } 4766 #endif 4767 4768 /* iterate through upper list, must be called under RCU read lock */ 4769 #define netdev_for_each_upper_dev_rcu(dev, updev, iter) \ 4770 for (iter = &(dev)->adj_list.upper, \ 4771 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \ 4772 updev; \ 4773 updev = netdev_upper_get_next_dev_rcu(dev, &(iter))) 4774 4775 int netdev_walk_all_upper_dev_rcu(struct net_device *dev, 4776 int (*fn)(struct net_device *upper_dev, 4777 struct netdev_nested_priv *priv), 4778 struct netdev_nested_priv *priv); 4779 4780 bool netdev_has_upper_dev_all_rcu(struct net_device *dev, 4781 struct net_device *upper_dev); 4782 4783 bool netdev_has_any_upper_dev(struct net_device *dev); 4784 4785 void *netdev_lower_get_next_private(struct net_device *dev, 4786 struct list_head **iter); 4787 void *netdev_lower_get_next_private_rcu(struct net_device *dev, 4788 struct list_head **iter); 4789 4790 #define netdev_for_each_lower_private(dev, priv, iter) \ 4791 for (iter = (dev)->adj_list.lower.next, \ 4792 priv = netdev_lower_get_next_private(dev, &(iter)); \ 4793 priv; \ 4794 priv = netdev_lower_get_next_private(dev, &(iter))) 4795 4796 #define netdev_for_each_lower_private_rcu(dev, priv, iter) \ 4797 for (iter = &(dev)->adj_list.lower, \ 4798 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \ 4799 priv; \ 4800 priv = netdev_lower_get_next_private_rcu(dev, &(iter))) 4801 4802 void *netdev_lower_get_next(struct net_device *dev, 4803 struct list_head **iter); 4804 4805 #define netdev_for_each_lower_dev(dev, ldev, iter) \ 4806 for (iter = (dev)->adj_list.lower.next, \ 4807 ldev = netdev_lower_get_next(dev, &(iter)); \ 4808 ldev; \ 4809 ldev = netdev_lower_get_next(dev, &(iter))) 4810 4811 struct net_device *netdev_next_lower_dev_rcu(struct net_device *dev, 4812 struct list_head **iter); 4813 int netdev_walk_all_lower_dev(struct net_device *dev, 4814 int (*fn)(struct net_device *lower_dev, 4815 struct netdev_nested_priv *priv), 4816 struct netdev_nested_priv *priv); 4817 int netdev_walk_all_lower_dev_rcu(struct net_device *dev, 4818 int (*fn)(struct net_device *lower_dev, 4819 struct netdev_nested_priv *priv), 4820 struct netdev_nested_priv *priv); 4821 4822 void *netdev_adjacent_get_private(struct list_head *adj_list); 4823 void *netdev_lower_get_first_private_rcu(struct net_device *dev); 4824 struct net_device *netdev_master_upper_dev_get(struct net_device *dev); 4825 struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev); 4826 int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev, 4827 struct netlink_ext_ack *extack); 4828 int netdev_master_upper_dev_link(struct net_device *dev, 4829 struct net_device *upper_dev, 4830 void *upper_priv, void *upper_info, 4831 struct netlink_ext_ack *extack); 4832 void netdev_upper_dev_unlink(struct net_device *dev, 4833 struct net_device *upper_dev); 4834 int netdev_adjacent_change_prepare(struct net_device *old_dev, 4835 struct net_device *new_dev, 4836 struct net_device *dev, 4837 struct netlink_ext_ack *extack); 4838 void netdev_adjacent_change_commit(struct net_device *old_dev, 4839 struct net_device *new_dev, 4840 struct net_device *dev); 4841 void netdev_adjacent_change_abort(struct net_device *old_dev, 4842 struct net_device *new_dev, 4843 struct net_device *dev); 4844 void netdev_adjacent_rename_links(struct net_device *dev, char *oldname); 4845 void *netdev_lower_dev_get_private(struct net_device *dev, 4846 struct net_device *lower_dev); 4847 void netdev_lower_state_changed(struct net_device *lower_dev, 4848 void *lower_state_info); 4849 4850 /* RSS keys are 40 or 52 bytes long */ 4851 #define NETDEV_RSS_KEY_LEN 52 4852 extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN] __read_mostly; 4853 void netdev_rss_key_fill(void *buffer, size_t len); 4854 4855 int skb_checksum_help(struct sk_buff *skb); 4856 int skb_crc32c_csum_help(struct sk_buff *skb); 4857 int skb_csum_hwoffload_help(struct sk_buff *skb, 4858 const netdev_features_t features); 4859 4860 struct sk_buff *__skb_gso_segment(struct sk_buff *skb, 4861 netdev_features_t features, bool tx_path); 4862 struct sk_buff *skb_mac_gso_segment(struct sk_buff *skb, 4863 netdev_features_t features); 4864 4865 struct netdev_bonding_info { 4866 ifslave slave; 4867 ifbond master; 4868 }; 4869 4870 struct netdev_notifier_bonding_info { 4871 struct netdev_notifier_info info; /* must be first */ 4872 struct netdev_bonding_info bonding_info; 4873 }; 4874 4875 void netdev_bonding_info_change(struct net_device *dev, 4876 struct netdev_bonding_info *bonding_info); 4877 4878 #if IS_ENABLED(CONFIG_ETHTOOL_NETLINK) 4879 void ethtool_notify(struct net_device *dev, unsigned int cmd, const void *data); 4880 #else 4881 static inline void ethtool_notify(struct net_device *dev, unsigned int cmd, 4882 const void *data) 4883 { 4884 } 4885 #endif 4886 4887 static inline 4888 struct sk_buff *skb_gso_segment(struct sk_buff *skb, netdev_features_t features) 4889 { 4890 return __skb_gso_segment(skb, features, true); 4891 } 4892 __be16 skb_network_protocol(struct sk_buff *skb, int *depth); 4893 4894 static inline bool can_checksum_protocol(netdev_features_t features, 4895 __be16 protocol) 4896 { 4897 if (protocol == htons(ETH_P_FCOE)) 4898 return !!(features & NETIF_F_FCOE_CRC); 4899 4900 /* Assume this is an IP checksum (not SCTP CRC) */ 4901 4902 if (features & NETIF_F_HW_CSUM) { 4903 /* Can checksum everything */ 4904 return true; 4905 } 4906 4907 switch (protocol) { 4908 case htons(ETH_P_IP): 4909 return !!(features & NETIF_F_IP_CSUM); 4910 case htons(ETH_P_IPV6): 4911 return !!(features & NETIF_F_IPV6_CSUM); 4912 default: 4913 return false; 4914 } 4915 } 4916 4917 #ifdef CONFIG_BUG 4918 void netdev_rx_csum_fault(struct net_device *dev, struct sk_buff *skb); 4919 #else 4920 static inline void netdev_rx_csum_fault(struct net_device *dev, 4921 struct sk_buff *skb) 4922 { 4923 } 4924 #endif 4925 /* rx skb timestamps */ 4926 void net_enable_timestamp(void); 4927 void net_disable_timestamp(void); 4928 4929 #ifdef CONFIG_PROC_FS 4930 int __init dev_proc_init(void); 4931 #else 4932 #define dev_proc_init() 0 4933 #endif 4934 4935 static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops, 4936 struct sk_buff *skb, struct net_device *dev, 4937 bool more) 4938 { 4939 __this_cpu_write(softnet_data.xmit.more, more); 4940 return ops->ndo_start_xmit(skb, dev); 4941 } 4942 4943 static inline bool netdev_xmit_more(void) 4944 { 4945 return __this_cpu_read(softnet_data.xmit.more); 4946 } 4947 4948 static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev, 4949 struct netdev_queue *txq, bool more) 4950 { 4951 const struct net_device_ops *ops = dev->netdev_ops; 4952 netdev_tx_t rc; 4953 4954 rc = __netdev_start_xmit(ops, skb, dev, more); 4955 if (rc == NETDEV_TX_OK) 4956 txq_trans_update(txq); 4957 4958 return rc; 4959 } 4960 4961 int netdev_class_create_file_ns(const struct class_attribute *class_attr, 4962 const void *ns); 4963 void netdev_class_remove_file_ns(const struct class_attribute *class_attr, 4964 const void *ns); 4965 4966 extern const struct kobj_ns_type_operations net_ns_type_operations; 4967 4968 const char *netdev_drivername(const struct net_device *dev); 4969 4970 void linkwatch_run_queue(void); 4971 4972 static inline netdev_features_t netdev_intersect_features(netdev_features_t f1, 4973 netdev_features_t f2) 4974 { 4975 if ((f1 ^ f2) & NETIF_F_HW_CSUM) { 4976 if (f1 & NETIF_F_HW_CSUM) 4977 f1 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM); 4978 else 4979 f2 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM); 4980 } 4981 4982 return f1 & f2; 4983 } 4984 4985 static inline netdev_features_t netdev_get_wanted_features( 4986 struct net_device *dev) 4987 { 4988 return (dev->features & ~dev->hw_features) | dev->wanted_features; 4989 } 4990 netdev_features_t netdev_increment_features(netdev_features_t all, 4991 netdev_features_t one, netdev_features_t mask); 4992 4993 /* Allow TSO being used on stacked device : 4994 * Performing the GSO segmentation before last device 4995 * is a performance improvement. 4996 */ 4997 static inline netdev_features_t netdev_add_tso_features(netdev_features_t features, 4998 netdev_features_t mask) 4999 { 5000 return netdev_increment_features(features, NETIF_F_ALL_TSO, mask); 5001 } 5002 5003 int __netdev_update_features(struct net_device *dev); 5004 void netdev_update_features(struct net_device *dev); 5005 void netdev_change_features(struct net_device *dev); 5006 5007 void netif_stacked_transfer_operstate(const struct net_device *rootdev, 5008 struct net_device *dev); 5009 5010 netdev_features_t passthru_features_check(struct sk_buff *skb, 5011 struct net_device *dev, 5012 netdev_features_t features); 5013 netdev_features_t netif_skb_features(struct sk_buff *skb); 5014 5015 static inline bool net_gso_ok(netdev_features_t features, int gso_type) 5016 { 5017 netdev_features_t feature = (netdev_features_t)gso_type << NETIF_F_GSO_SHIFT; 5018 5019 /* check flags correspondence */ 5020 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT)); 5021 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT)); 5022 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT)); 5023 BUILD_BUG_ON(SKB_GSO_TCP_FIXEDID != (NETIF_F_TSO_MANGLEID >> NETIF_F_GSO_SHIFT)); 5024 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT)); 5025 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT)); 5026 BUILD_BUG_ON(SKB_GSO_GRE != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT)); 5027 BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT)); 5028 BUILD_BUG_ON(SKB_GSO_IPXIP4 != (NETIF_F_GSO_IPXIP4 >> NETIF_F_GSO_SHIFT)); 5029 BUILD_BUG_ON(SKB_GSO_IPXIP6 != (NETIF_F_GSO_IPXIP6 >> NETIF_F_GSO_SHIFT)); 5030 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT)); 5031 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT)); 5032 BUILD_BUG_ON(SKB_GSO_PARTIAL != (NETIF_F_GSO_PARTIAL >> NETIF_F_GSO_SHIFT)); 5033 BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT)); 5034 BUILD_BUG_ON(SKB_GSO_SCTP != (NETIF_F_GSO_SCTP >> NETIF_F_GSO_SHIFT)); 5035 BUILD_BUG_ON(SKB_GSO_ESP != (NETIF_F_GSO_ESP >> NETIF_F_GSO_SHIFT)); 5036 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_GSO_UDP >> NETIF_F_GSO_SHIFT)); 5037 BUILD_BUG_ON(SKB_GSO_UDP_L4 != (NETIF_F_GSO_UDP_L4 >> NETIF_F_GSO_SHIFT)); 5038 BUILD_BUG_ON(SKB_GSO_FRAGLIST != (NETIF_F_GSO_FRAGLIST >> NETIF_F_GSO_SHIFT)); 5039 5040 return (features & feature) == feature; 5041 } 5042 5043 static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features) 5044 { 5045 return net_gso_ok(features, skb_shinfo(skb)->gso_type) && 5046 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST)); 5047 } 5048 5049 static inline bool netif_needs_gso(struct sk_buff *skb, 5050 netdev_features_t features) 5051 { 5052 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) || 5053 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) && 5054 (skb->ip_summed != CHECKSUM_UNNECESSARY))); 5055 } 5056 5057 static inline void netif_set_gso_max_size(struct net_device *dev, 5058 unsigned int size) 5059 { 5060 dev->gso_max_size = size; 5061 } 5062 5063 static inline void skb_gso_error_unwind(struct sk_buff *skb, __be16 protocol, 5064 int pulled_hlen, u16 mac_offset, 5065 int mac_len) 5066 { 5067 skb->protocol = protocol; 5068 skb->encapsulation = 1; 5069 skb_push(skb, pulled_hlen); 5070 skb_reset_transport_header(skb); 5071 skb->mac_header = mac_offset; 5072 skb->network_header = skb->mac_header + mac_len; 5073 skb->mac_len = mac_len; 5074 } 5075 5076 static inline bool netif_is_macsec(const struct net_device *dev) 5077 { 5078 return dev->priv_flags & IFF_MACSEC; 5079 } 5080 5081 static inline bool netif_is_macvlan(const struct net_device *dev) 5082 { 5083 return dev->priv_flags & IFF_MACVLAN; 5084 } 5085 5086 static inline bool netif_is_macvlan_port(const struct net_device *dev) 5087 { 5088 return dev->priv_flags & IFF_MACVLAN_PORT; 5089 } 5090 5091 static inline bool netif_is_bond_master(const struct net_device *dev) 5092 { 5093 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING; 5094 } 5095 5096 static inline bool netif_is_bond_slave(const struct net_device *dev) 5097 { 5098 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING; 5099 } 5100 5101 static inline bool netif_supports_nofcs(struct net_device *dev) 5102 { 5103 return dev->priv_flags & IFF_SUPP_NOFCS; 5104 } 5105 5106 static inline bool netif_has_l3_rx_handler(const struct net_device *dev) 5107 { 5108 return dev->priv_flags & IFF_L3MDEV_RX_HANDLER; 5109 } 5110 5111 static inline bool netif_is_l3_master(const struct net_device *dev) 5112 { 5113 return dev->priv_flags & IFF_L3MDEV_MASTER; 5114 } 5115 5116 static inline bool netif_is_l3_slave(const struct net_device *dev) 5117 { 5118 return dev->priv_flags & IFF_L3MDEV_SLAVE; 5119 } 5120 5121 static inline bool netif_is_bridge_master(const struct net_device *dev) 5122 { 5123 return dev->priv_flags & IFF_EBRIDGE; 5124 } 5125 5126 static inline bool netif_is_bridge_port(const struct net_device *dev) 5127 { 5128 return dev->priv_flags & IFF_BRIDGE_PORT; 5129 } 5130 5131 static inline bool netif_is_ovs_master(const struct net_device *dev) 5132 { 5133 return dev->priv_flags & IFF_OPENVSWITCH; 5134 } 5135 5136 static inline bool netif_is_ovs_port(const struct net_device *dev) 5137 { 5138 return dev->priv_flags & IFF_OVS_DATAPATH; 5139 } 5140 5141 static inline bool netif_is_any_bridge_port(const struct net_device *dev) 5142 { 5143 return netif_is_bridge_port(dev) || netif_is_ovs_port(dev); 5144 } 5145 5146 static inline bool netif_is_team_master(const struct net_device *dev) 5147 { 5148 return dev->priv_flags & IFF_TEAM; 5149 } 5150 5151 static inline bool netif_is_team_port(const struct net_device *dev) 5152 { 5153 return dev->priv_flags & IFF_TEAM_PORT; 5154 } 5155 5156 static inline bool netif_is_lag_master(const struct net_device *dev) 5157 { 5158 return netif_is_bond_master(dev) || netif_is_team_master(dev); 5159 } 5160 5161 static inline bool netif_is_lag_port(const struct net_device *dev) 5162 { 5163 return netif_is_bond_slave(dev) || netif_is_team_port(dev); 5164 } 5165 5166 static inline bool netif_is_rxfh_configured(const struct net_device *dev) 5167 { 5168 return dev->priv_flags & IFF_RXFH_CONFIGURED; 5169 } 5170 5171 static inline bool netif_is_failover(const struct net_device *dev) 5172 { 5173 return dev->priv_flags & IFF_FAILOVER; 5174 } 5175 5176 static inline bool netif_is_failover_slave(const struct net_device *dev) 5177 { 5178 return dev->priv_flags & IFF_FAILOVER_SLAVE; 5179 } 5180 5181 /* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */ 5182 static inline void netif_keep_dst(struct net_device *dev) 5183 { 5184 dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM); 5185 } 5186 5187 /* return true if dev can't cope with mtu frames that need vlan tag insertion */ 5188 static inline bool netif_reduces_vlan_mtu(struct net_device *dev) 5189 { 5190 /* TODO: reserve and use an additional IFF bit, if we get more users */ 5191 return dev->priv_flags & IFF_MACSEC; 5192 } 5193 5194 extern struct pernet_operations __net_initdata loopback_net_ops; 5195 5196 /* Logging, debugging and troubleshooting/diagnostic helpers. */ 5197 5198 /* netdev_printk helpers, similar to dev_printk */ 5199 5200 static inline const char *netdev_name(const struct net_device *dev) 5201 { 5202 if (!dev->name[0] || strchr(dev->name, '%')) 5203 return "(unnamed net_device)"; 5204 return dev->name; 5205 } 5206 5207 static inline bool netdev_unregistering(const struct net_device *dev) 5208 { 5209 return dev->reg_state == NETREG_UNREGISTERING; 5210 } 5211 5212 static inline const char *netdev_reg_state(const struct net_device *dev) 5213 { 5214 switch (dev->reg_state) { 5215 case NETREG_UNINITIALIZED: return " (uninitialized)"; 5216 case NETREG_REGISTERED: return ""; 5217 case NETREG_UNREGISTERING: return " (unregistering)"; 5218 case NETREG_UNREGISTERED: return " (unregistered)"; 5219 case NETREG_RELEASED: return " (released)"; 5220 case NETREG_DUMMY: return " (dummy)"; 5221 } 5222 5223 WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, dev->reg_state); 5224 return " (unknown)"; 5225 } 5226 5227 __printf(3, 4) __cold 5228 void netdev_printk(const char *level, const struct net_device *dev, 5229 const char *format, ...); 5230 __printf(2, 3) __cold 5231 void netdev_emerg(const struct net_device *dev, const char *format, ...); 5232 __printf(2, 3) __cold 5233 void netdev_alert(const struct net_device *dev, const char *format, ...); 5234 __printf(2, 3) __cold 5235 void netdev_crit(const struct net_device *dev, const char *format, ...); 5236 __printf(2, 3) __cold 5237 void netdev_err(const struct net_device *dev, const char *format, ...); 5238 __printf(2, 3) __cold 5239 void netdev_warn(const struct net_device *dev, const char *format, ...); 5240 __printf(2, 3) __cold 5241 void netdev_notice(const struct net_device *dev, const char *format, ...); 5242 __printf(2, 3) __cold 5243 void netdev_info(const struct net_device *dev, const char *format, ...); 5244 5245 #define netdev_level_once(level, dev, fmt, ...) \ 5246 do { \ 5247 static bool __print_once __read_mostly; \ 5248 \ 5249 if (!__print_once) { \ 5250 __print_once = true; \ 5251 netdev_printk(level, dev, fmt, ##__VA_ARGS__); \ 5252 } \ 5253 } while (0) 5254 5255 #define netdev_emerg_once(dev, fmt, ...) \ 5256 netdev_level_once(KERN_EMERG, dev, fmt, ##__VA_ARGS__) 5257 #define netdev_alert_once(dev, fmt, ...) \ 5258 netdev_level_once(KERN_ALERT, dev, fmt, ##__VA_ARGS__) 5259 #define netdev_crit_once(dev, fmt, ...) \ 5260 netdev_level_once(KERN_CRIT, dev, fmt, ##__VA_ARGS__) 5261 #define netdev_err_once(dev, fmt, ...) \ 5262 netdev_level_once(KERN_ERR, dev, fmt, ##__VA_ARGS__) 5263 #define netdev_warn_once(dev, fmt, ...) \ 5264 netdev_level_once(KERN_WARNING, dev, fmt, ##__VA_ARGS__) 5265 #define netdev_notice_once(dev, fmt, ...) \ 5266 netdev_level_once(KERN_NOTICE, dev, fmt, ##__VA_ARGS__) 5267 #define netdev_info_once(dev, fmt, ...) \ 5268 netdev_level_once(KERN_INFO, dev, fmt, ##__VA_ARGS__) 5269 5270 #define MODULE_ALIAS_NETDEV(device) \ 5271 MODULE_ALIAS("netdev-" device) 5272 5273 #if defined(CONFIG_DYNAMIC_DEBUG) || \ 5274 (defined(CONFIG_DYNAMIC_DEBUG_CORE) && defined(DYNAMIC_DEBUG_MODULE)) 5275 #define netdev_dbg(__dev, format, args...) \ 5276 do { \ 5277 dynamic_netdev_dbg(__dev, format, ##args); \ 5278 } while (0) 5279 #elif defined(DEBUG) 5280 #define netdev_dbg(__dev, format, args...) \ 5281 netdev_printk(KERN_DEBUG, __dev, format, ##args) 5282 #else 5283 #define netdev_dbg(__dev, format, args...) \ 5284 ({ \ 5285 if (0) \ 5286 netdev_printk(KERN_DEBUG, __dev, format, ##args); \ 5287 }) 5288 #endif 5289 5290 #if defined(VERBOSE_DEBUG) 5291 #define netdev_vdbg netdev_dbg 5292 #else 5293 5294 #define netdev_vdbg(dev, format, args...) \ 5295 ({ \ 5296 if (0) \ 5297 netdev_printk(KERN_DEBUG, dev, format, ##args); \ 5298 0; \ 5299 }) 5300 #endif 5301 5302 /* 5303 * netdev_WARN() acts like dev_printk(), but with the key difference 5304 * of using a WARN/WARN_ON to get the message out, including the 5305 * file/line information and a backtrace. 5306 */ 5307 #define netdev_WARN(dev, format, args...) \ 5308 WARN(1, "netdevice: %s%s: " format, netdev_name(dev), \ 5309 netdev_reg_state(dev), ##args) 5310 5311 #define netdev_WARN_ONCE(dev, format, args...) \ 5312 WARN_ONCE(1, "netdevice: %s%s: " format, netdev_name(dev), \ 5313 netdev_reg_state(dev), ##args) 5314 5315 /* netif printk helpers, similar to netdev_printk */ 5316 5317 #define netif_printk(priv, type, level, dev, fmt, args...) \ 5318 do { \ 5319 if (netif_msg_##type(priv)) \ 5320 netdev_printk(level, (dev), fmt, ##args); \ 5321 } while (0) 5322 5323 #define netif_level(level, priv, type, dev, fmt, args...) \ 5324 do { \ 5325 if (netif_msg_##type(priv)) \ 5326 netdev_##level(dev, fmt, ##args); \ 5327 } while (0) 5328 5329 #define netif_emerg(priv, type, dev, fmt, args...) \ 5330 netif_level(emerg, priv, type, dev, fmt, ##args) 5331 #define netif_alert(priv, type, dev, fmt, args...) \ 5332 netif_level(alert, priv, type, dev, fmt, ##args) 5333 #define netif_crit(priv, type, dev, fmt, args...) \ 5334 netif_level(crit, priv, type, dev, fmt, ##args) 5335 #define netif_err(priv, type, dev, fmt, args...) \ 5336 netif_level(err, priv, type, dev, fmt, ##args) 5337 #define netif_warn(priv, type, dev, fmt, args...) \ 5338 netif_level(warn, priv, type, dev, fmt, ##args) 5339 #define netif_notice(priv, type, dev, fmt, args...) \ 5340 netif_level(notice, priv, type, dev, fmt, ##args) 5341 #define netif_info(priv, type, dev, fmt, args...) \ 5342 netif_level(info, priv, type, dev, fmt, ##args) 5343 5344 #if defined(CONFIG_DYNAMIC_DEBUG) || \ 5345 (defined(CONFIG_DYNAMIC_DEBUG_CORE) && defined(DYNAMIC_DEBUG_MODULE)) 5346 #define netif_dbg(priv, type, netdev, format, args...) \ 5347 do { \ 5348 if (netif_msg_##type(priv)) \ 5349 dynamic_netdev_dbg(netdev, format, ##args); \ 5350 } while (0) 5351 #elif defined(DEBUG) 5352 #define netif_dbg(priv, type, dev, format, args...) \ 5353 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args) 5354 #else 5355 #define netif_dbg(priv, type, dev, format, args...) \ 5356 ({ \ 5357 if (0) \ 5358 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \ 5359 0; \ 5360 }) 5361 #endif 5362 5363 /* if @cond then downgrade to debug, else print at @level */ 5364 #define netif_cond_dbg(priv, type, netdev, cond, level, fmt, args...) \ 5365 do { \ 5366 if (cond) \ 5367 netif_dbg(priv, type, netdev, fmt, ##args); \ 5368 else \ 5369 netif_ ## level(priv, type, netdev, fmt, ##args); \ 5370 } while (0) 5371 5372 #if defined(VERBOSE_DEBUG) 5373 #define netif_vdbg netif_dbg 5374 #else 5375 #define netif_vdbg(priv, type, dev, format, args...) \ 5376 ({ \ 5377 if (0) \ 5378 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \ 5379 0; \ 5380 }) 5381 #endif 5382 5383 /* 5384 * The list of packet types we will receive (as opposed to discard) 5385 * and the routines to invoke. 5386 * 5387 * Why 16. Because with 16 the only overlap we get on a hash of the 5388 * low nibble of the protocol value is RARP/SNAP/X.25. 5389 * 5390 * 0800 IP 5391 * 0001 802.3 5392 * 0002 AX.25 5393 * 0004 802.2 5394 * 8035 RARP 5395 * 0005 SNAP 5396 * 0805 X.25 5397 * 0806 ARP 5398 * 8137 IPX 5399 * 0009 Localtalk 5400 * 86DD IPv6 5401 */ 5402 #define PTYPE_HASH_SIZE (16) 5403 #define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1) 5404 5405 extern struct list_head ptype_all __read_mostly; 5406 extern struct list_head ptype_base[PTYPE_HASH_SIZE] __read_mostly; 5407 5408 extern struct net_device *blackhole_netdev; 5409 5410 #endif /* _LINUX_NETDEVICE_H */ 5411