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 * Old-style ioctl entry point. This is used internally by the 1090 * appletalk and ieee802154 subsystems but is no longer called by 1091 * the device ioctl handler. 1092 * 1093 * int (*ndo_siocbond)(struct net_device *dev, struct ifreq *ifr, int cmd); 1094 * Used by the bonding driver for its device specific ioctls: 1095 * SIOCBONDENSLAVE, SIOCBONDRELEASE, SIOCBONDSETHWADDR, SIOCBONDCHANGEACTIVE, 1096 * SIOCBONDSLAVEINFOQUERY, and SIOCBONDINFOQUERY 1097 * 1098 * * int (*ndo_eth_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd); 1099 * Called for ethernet specific ioctls: SIOCGMIIPHY, SIOCGMIIREG, 1100 * SIOCSMIIREG, SIOCSHWTSTAMP and SIOCGHWTSTAMP. 1101 * 1102 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map); 1103 * Used to set network devices bus interface parameters. This interface 1104 * is retained for legacy reasons; new devices should use the bus 1105 * interface (PCI) for low level management. 1106 * 1107 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu); 1108 * Called when a user wants to change the Maximum Transfer Unit 1109 * of a device. 1110 * 1111 * void (*ndo_tx_timeout)(struct net_device *dev, unsigned int txqueue); 1112 * Callback used when the transmitter has not made any progress 1113 * for dev->watchdog ticks. 1114 * 1115 * void (*ndo_get_stats64)(struct net_device *dev, 1116 * struct rtnl_link_stats64 *storage); 1117 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev); 1118 * Called when a user wants to get the network device usage 1119 * statistics. Drivers must do one of the following: 1120 * 1. Define @ndo_get_stats64 to fill in a zero-initialised 1121 * rtnl_link_stats64 structure passed by the caller. 1122 * 2. Define @ndo_get_stats to update a net_device_stats structure 1123 * (which should normally be dev->stats) and return a pointer to 1124 * it. The structure may be changed asynchronously only if each 1125 * field is written atomically. 1126 * 3. Update dev->stats asynchronously and atomically, and define 1127 * neither operation. 1128 * 1129 * bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id) 1130 * Return true if this device supports offload stats of this attr_id. 1131 * 1132 * int (*ndo_get_offload_stats)(int attr_id, const struct net_device *dev, 1133 * void *attr_data) 1134 * Get statistics for offload operations by attr_id. Write it into the 1135 * attr_data pointer. 1136 * 1137 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16 vid); 1138 * If device supports VLAN filtering this function is called when a 1139 * VLAN id is registered. 1140 * 1141 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, __be16 proto, u16 vid); 1142 * If device supports VLAN filtering this function is called when a 1143 * VLAN id is unregistered. 1144 * 1145 * void (*ndo_poll_controller)(struct net_device *dev); 1146 * 1147 * SR-IOV management functions. 1148 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac); 1149 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan, 1150 * u8 qos, __be16 proto); 1151 * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate, 1152 * int max_tx_rate); 1153 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting); 1154 * int (*ndo_set_vf_trust)(struct net_device *dev, int vf, bool setting); 1155 * int (*ndo_get_vf_config)(struct net_device *dev, 1156 * int vf, struct ifla_vf_info *ivf); 1157 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state); 1158 * int (*ndo_set_vf_port)(struct net_device *dev, int vf, 1159 * struct nlattr *port[]); 1160 * 1161 * Enable or disable the VF ability to query its RSS Redirection Table and 1162 * Hash Key. This is needed since on some devices VF share this information 1163 * with PF and querying it may introduce a theoretical security risk. 1164 * int (*ndo_set_vf_rss_query_en)(struct net_device *dev, int vf, bool setting); 1165 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb); 1166 * int (*ndo_setup_tc)(struct net_device *dev, enum tc_setup_type type, 1167 * void *type_data); 1168 * Called to setup any 'tc' scheduler, classifier or action on @dev. 1169 * This is always called from the stack with the rtnl lock held and netif 1170 * tx queues stopped. This allows the netdevice to perform queue 1171 * management safely. 1172 * 1173 * Fiber Channel over Ethernet (FCoE) offload functions. 1174 * int (*ndo_fcoe_enable)(struct net_device *dev); 1175 * Called when the FCoE protocol stack wants to start using LLD for FCoE 1176 * so the underlying device can perform whatever needed configuration or 1177 * initialization to support acceleration of FCoE traffic. 1178 * 1179 * int (*ndo_fcoe_disable)(struct net_device *dev); 1180 * Called when the FCoE protocol stack wants to stop using LLD for FCoE 1181 * so the underlying device can perform whatever needed clean-ups to 1182 * stop supporting acceleration of FCoE traffic. 1183 * 1184 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid, 1185 * struct scatterlist *sgl, unsigned int sgc); 1186 * Called when the FCoE Initiator wants to initialize an I/O that 1187 * is a possible candidate for Direct Data Placement (DDP). The LLD can 1188 * perform necessary setup and returns 1 to indicate the device is set up 1189 * successfully to perform DDP on this I/O, otherwise this returns 0. 1190 * 1191 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid); 1192 * Called when the FCoE Initiator/Target is done with the DDPed I/O as 1193 * indicated by the FC exchange id 'xid', so the underlying device can 1194 * clean up and reuse resources for later DDP requests. 1195 * 1196 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid, 1197 * struct scatterlist *sgl, unsigned int sgc); 1198 * Called when the FCoE Target wants to initialize an I/O that 1199 * is a possible candidate for Direct Data Placement (DDP). The LLD can 1200 * perform necessary setup and returns 1 to indicate the device is set up 1201 * successfully to perform DDP on this I/O, otherwise this returns 0. 1202 * 1203 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev, 1204 * struct netdev_fcoe_hbainfo *hbainfo); 1205 * Called when the FCoE Protocol stack wants information on the underlying 1206 * device. This information is utilized by the FCoE protocol stack to 1207 * register attributes with Fiber Channel management service as per the 1208 * FC-GS Fabric Device Management Information(FDMI) specification. 1209 * 1210 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type); 1211 * Called when the underlying device wants to override default World Wide 1212 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own 1213 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE 1214 * protocol stack to use. 1215 * 1216 * RFS acceleration. 1217 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb, 1218 * u16 rxq_index, u32 flow_id); 1219 * Set hardware filter for RFS. rxq_index is the target queue index; 1220 * flow_id is a flow ID to be passed to rps_may_expire_flow() later. 1221 * Return the filter ID on success, or a negative error code. 1222 * 1223 * Slave management functions (for bridge, bonding, etc). 1224 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev); 1225 * Called to make another netdev an underling. 1226 * 1227 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev); 1228 * Called to release previously enslaved netdev. 1229 * 1230 * struct net_device *(*ndo_get_xmit_slave)(struct net_device *dev, 1231 * struct sk_buff *skb, 1232 * bool all_slaves); 1233 * Get the xmit slave of master device. If all_slaves is true, function 1234 * assume all the slaves can transmit. 1235 * 1236 * Feature/offload setting functions. 1237 * netdev_features_t (*ndo_fix_features)(struct net_device *dev, 1238 * netdev_features_t features); 1239 * Adjusts the requested feature flags according to device-specific 1240 * constraints, and returns the resulting flags. Must not modify 1241 * the device state. 1242 * 1243 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features); 1244 * Called to update device configuration to new features. Passed 1245 * feature set might be less than what was returned by ndo_fix_features()). 1246 * Must return >0 or -errno if it changed dev->features itself. 1247 * 1248 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[], 1249 * struct net_device *dev, 1250 * const unsigned char *addr, u16 vid, u16 flags, 1251 * struct netlink_ext_ack *extack); 1252 * Adds an FDB entry to dev for addr. 1253 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[], 1254 * struct net_device *dev, 1255 * const unsigned char *addr, u16 vid) 1256 * Deletes the FDB entry from dev coresponding to addr. 1257 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb, 1258 * struct net_device *dev, struct net_device *filter_dev, 1259 * int *idx) 1260 * Used to add FDB entries to dump requests. Implementers should add 1261 * entries to skb and update idx with the number of entries. 1262 * 1263 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh, 1264 * u16 flags, struct netlink_ext_ack *extack) 1265 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq, 1266 * struct net_device *dev, u32 filter_mask, 1267 * int nlflags) 1268 * int (*ndo_bridge_dellink)(struct net_device *dev, struct nlmsghdr *nlh, 1269 * u16 flags); 1270 * 1271 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier); 1272 * Called to change device carrier. Soft-devices (like dummy, team, etc) 1273 * which do not represent real hardware may define this to allow their 1274 * userspace components to manage their virtual carrier state. Devices 1275 * that determine carrier state from physical hardware properties (eg 1276 * network cables) or protocol-dependent mechanisms (eg 1277 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function. 1278 * 1279 * int (*ndo_get_phys_port_id)(struct net_device *dev, 1280 * struct netdev_phys_item_id *ppid); 1281 * Called to get ID of physical port of this device. If driver does 1282 * not implement this, it is assumed that the hw is not able to have 1283 * multiple net devices on single physical port. 1284 * 1285 * int (*ndo_get_port_parent_id)(struct net_device *dev, 1286 * struct netdev_phys_item_id *ppid) 1287 * Called to get the parent ID of the physical port of this device. 1288 * 1289 * void* (*ndo_dfwd_add_station)(struct net_device *pdev, 1290 * struct net_device *dev) 1291 * Called by upper layer devices to accelerate switching or other 1292 * station functionality into hardware. 'pdev is the lowerdev 1293 * to use for the offload and 'dev' is the net device that will 1294 * back the offload. Returns a pointer to the private structure 1295 * the upper layer will maintain. 1296 * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv) 1297 * Called by upper layer device to delete the station created 1298 * by 'ndo_dfwd_add_station'. 'pdev' is the net device backing 1299 * the station and priv is the structure returned by the add 1300 * operation. 1301 * int (*ndo_set_tx_maxrate)(struct net_device *dev, 1302 * int queue_index, u32 maxrate); 1303 * Called when a user wants to set a max-rate limitation of specific 1304 * TX queue. 1305 * int (*ndo_get_iflink)(const struct net_device *dev); 1306 * Called to get the iflink value of this device. 1307 * void (*ndo_change_proto_down)(struct net_device *dev, 1308 * bool proto_down); 1309 * This function is used to pass protocol port error state information 1310 * to the switch driver. The switch driver can react to the proto_down 1311 * by doing a phys down on the associated switch port. 1312 * int (*ndo_fill_metadata_dst)(struct net_device *dev, struct sk_buff *skb); 1313 * This function is used to get egress tunnel information for given skb. 1314 * This is useful for retrieving outer tunnel header parameters while 1315 * sampling packet. 1316 * void (*ndo_set_rx_headroom)(struct net_device *dev, int needed_headroom); 1317 * This function is used to specify the headroom that the skb must 1318 * consider when allocation skb during packet reception. Setting 1319 * appropriate rx headroom value allows avoiding skb head copy on 1320 * forward. Setting a negative value resets the rx headroom to the 1321 * default value. 1322 * int (*ndo_bpf)(struct net_device *dev, struct netdev_bpf *bpf); 1323 * This function is used to set or query state related to XDP on the 1324 * netdevice and manage BPF offload. See definition of 1325 * enum bpf_netdev_command for details. 1326 * int (*ndo_xdp_xmit)(struct net_device *dev, int n, struct xdp_frame **xdp, 1327 * u32 flags); 1328 * This function is used to submit @n XDP packets for transmit on a 1329 * netdevice. Returns number of frames successfully transmitted, frames 1330 * that got dropped are freed/returned via xdp_return_frame(). 1331 * Returns negative number, means general error invoking ndo, meaning 1332 * no frames were xmit'ed and core-caller will free all frames. 1333 * int (*ndo_xsk_wakeup)(struct net_device *dev, u32 queue_id, u32 flags); 1334 * This function is used to wake up the softirq, ksoftirqd or kthread 1335 * responsible for sending and/or receiving packets on a specific 1336 * queue id bound to an AF_XDP socket. The flags field specifies if 1337 * only RX, only Tx, or both should be woken up using the flags 1338 * XDP_WAKEUP_RX and XDP_WAKEUP_TX. 1339 * struct devlink_port *(*ndo_get_devlink_port)(struct net_device *dev); 1340 * Get devlink port instance associated with a given netdev. 1341 * Called with a reference on the netdevice and devlink locks only, 1342 * rtnl_lock is not held. 1343 * int (*ndo_tunnel_ctl)(struct net_device *dev, struct ip_tunnel_parm *p, 1344 * int cmd); 1345 * Add, change, delete or get information on an IPv4 tunnel. 1346 * struct net_device *(*ndo_get_peer_dev)(struct net_device *dev); 1347 * If a device is paired with a peer device, return the peer instance. 1348 * The caller must be under RCU read context. 1349 * int (*ndo_fill_forward_path)(struct net_device_path_ctx *ctx, struct net_device_path *path); 1350 * Get the forwarding path to reach the real device from the HW destination address 1351 */ 1352 struct net_device_ops { 1353 int (*ndo_init)(struct net_device *dev); 1354 void (*ndo_uninit)(struct net_device *dev); 1355 int (*ndo_open)(struct net_device *dev); 1356 int (*ndo_stop)(struct net_device *dev); 1357 netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb, 1358 struct net_device *dev); 1359 netdev_features_t (*ndo_features_check)(struct sk_buff *skb, 1360 struct net_device *dev, 1361 netdev_features_t features); 1362 u16 (*ndo_select_queue)(struct net_device *dev, 1363 struct sk_buff *skb, 1364 struct net_device *sb_dev); 1365 void (*ndo_change_rx_flags)(struct net_device *dev, 1366 int flags); 1367 void (*ndo_set_rx_mode)(struct net_device *dev); 1368 int (*ndo_set_mac_address)(struct net_device *dev, 1369 void *addr); 1370 int (*ndo_validate_addr)(struct net_device *dev); 1371 int (*ndo_do_ioctl)(struct net_device *dev, 1372 struct ifreq *ifr, int cmd); 1373 int (*ndo_eth_ioctl)(struct net_device *dev, 1374 struct ifreq *ifr, int cmd); 1375 int (*ndo_siocbond)(struct net_device *dev, 1376 struct ifreq *ifr, int cmd); 1377 int (*ndo_siocwandev)(struct net_device *dev, 1378 struct if_settings *ifs); 1379 int (*ndo_siocdevprivate)(struct net_device *dev, 1380 struct ifreq *ifr, 1381 void __user *data, int cmd); 1382 int (*ndo_set_config)(struct net_device *dev, 1383 struct ifmap *map); 1384 int (*ndo_change_mtu)(struct net_device *dev, 1385 int new_mtu); 1386 int (*ndo_neigh_setup)(struct net_device *dev, 1387 struct neigh_parms *); 1388 void (*ndo_tx_timeout) (struct net_device *dev, 1389 unsigned int txqueue); 1390 1391 void (*ndo_get_stats64)(struct net_device *dev, 1392 struct rtnl_link_stats64 *storage); 1393 bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id); 1394 int (*ndo_get_offload_stats)(int attr_id, 1395 const struct net_device *dev, 1396 void *attr_data); 1397 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev); 1398 1399 int (*ndo_vlan_rx_add_vid)(struct net_device *dev, 1400 __be16 proto, u16 vid); 1401 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, 1402 __be16 proto, u16 vid); 1403 #ifdef CONFIG_NET_POLL_CONTROLLER 1404 void (*ndo_poll_controller)(struct net_device *dev); 1405 int (*ndo_netpoll_setup)(struct net_device *dev, 1406 struct netpoll_info *info); 1407 void (*ndo_netpoll_cleanup)(struct net_device *dev); 1408 #endif 1409 int (*ndo_set_vf_mac)(struct net_device *dev, 1410 int queue, u8 *mac); 1411 int (*ndo_set_vf_vlan)(struct net_device *dev, 1412 int queue, u16 vlan, 1413 u8 qos, __be16 proto); 1414 int (*ndo_set_vf_rate)(struct net_device *dev, 1415 int vf, int min_tx_rate, 1416 int max_tx_rate); 1417 int (*ndo_set_vf_spoofchk)(struct net_device *dev, 1418 int vf, bool setting); 1419 int (*ndo_set_vf_trust)(struct net_device *dev, 1420 int vf, bool setting); 1421 int (*ndo_get_vf_config)(struct net_device *dev, 1422 int vf, 1423 struct ifla_vf_info *ivf); 1424 int (*ndo_set_vf_link_state)(struct net_device *dev, 1425 int vf, int link_state); 1426 int (*ndo_get_vf_stats)(struct net_device *dev, 1427 int vf, 1428 struct ifla_vf_stats 1429 *vf_stats); 1430 int (*ndo_set_vf_port)(struct net_device *dev, 1431 int vf, 1432 struct nlattr *port[]); 1433 int (*ndo_get_vf_port)(struct net_device *dev, 1434 int vf, struct sk_buff *skb); 1435 int (*ndo_get_vf_guid)(struct net_device *dev, 1436 int vf, 1437 struct ifla_vf_guid *node_guid, 1438 struct ifla_vf_guid *port_guid); 1439 int (*ndo_set_vf_guid)(struct net_device *dev, 1440 int vf, u64 guid, 1441 int guid_type); 1442 int (*ndo_set_vf_rss_query_en)( 1443 struct net_device *dev, 1444 int vf, bool setting); 1445 int (*ndo_setup_tc)(struct net_device *dev, 1446 enum tc_setup_type type, 1447 void *type_data); 1448 #if IS_ENABLED(CONFIG_FCOE) 1449 int (*ndo_fcoe_enable)(struct net_device *dev); 1450 int (*ndo_fcoe_disable)(struct net_device *dev); 1451 int (*ndo_fcoe_ddp_setup)(struct net_device *dev, 1452 u16 xid, 1453 struct scatterlist *sgl, 1454 unsigned int sgc); 1455 int (*ndo_fcoe_ddp_done)(struct net_device *dev, 1456 u16 xid); 1457 int (*ndo_fcoe_ddp_target)(struct net_device *dev, 1458 u16 xid, 1459 struct scatterlist *sgl, 1460 unsigned int sgc); 1461 int (*ndo_fcoe_get_hbainfo)(struct net_device *dev, 1462 struct netdev_fcoe_hbainfo *hbainfo); 1463 #endif 1464 1465 #if IS_ENABLED(CONFIG_LIBFCOE) 1466 #define NETDEV_FCOE_WWNN 0 1467 #define NETDEV_FCOE_WWPN 1 1468 int (*ndo_fcoe_get_wwn)(struct net_device *dev, 1469 u64 *wwn, int type); 1470 #endif 1471 1472 #ifdef CONFIG_RFS_ACCEL 1473 int (*ndo_rx_flow_steer)(struct net_device *dev, 1474 const struct sk_buff *skb, 1475 u16 rxq_index, 1476 u32 flow_id); 1477 #endif 1478 int (*ndo_add_slave)(struct net_device *dev, 1479 struct net_device *slave_dev, 1480 struct netlink_ext_ack *extack); 1481 int (*ndo_del_slave)(struct net_device *dev, 1482 struct net_device *slave_dev); 1483 struct net_device* (*ndo_get_xmit_slave)(struct net_device *dev, 1484 struct sk_buff *skb, 1485 bool all_slaves); 1486 struct net_device* (*ndo_sk_get_lower_dev)(struct net_device *dev, 1487 struct sock *sk); 1488 netdev_features_t (*ndo_fix_features)(struct net_device *dev, 1489 netdev_features_t features); 1490 int (*ndo_set_features)(struct net_device *dev, 1491 netdev_features_t features); 1492 int (*ndo_neigh_construct)(struct net_device *dev, 1493 struct neighbour *n); 1494 void (*ndo_neigh_destroy)(struct net_device *dev, 1495 struct neighbour *n); 1496 1497 int (*ndo_fdb_add)(struct ndmsg *ndm, 1498 struct nlattr *tb[], 1499 struct net_device *dev, 1500 const unsigned char *addr, 1501 u16 vid, 1502 u16 flags, 1503 struct netlink_ext_ack *extack); 1504 int (*ndo_fdb_del)(struct ndmsg *ndm, 1505 struct nlattr *tb[], 1506 struct net_device *dev, 1507 const unsigned char *addr, 1508 u16 vid); 1509 int (*ndo_fdb_dump)(struct sk_buff *skb, 1510 struct netlink_callback *cb, 1511 struct net_device *dev, 1512 struct net_device *filter_dev, 1513 int *idx); 1514 int (*ndo_fdb_get)(struct sk_buff *skb, 1515 struct nlattr *tb[], 1516 struct net_device *dev, 1517 const unsigned char *addr, 1518 u16 vid, u32 portid, u32 seq, 1519 struct netlink_ext_ack *extack); 1520 int (*ndo_bridge_setlink)(struct net_device *dev, 1521 struct nlmsghdr *nlh, 1522 u16 flags, 1523 struct netlink_ext_ack *extack); 1524 int (*ndo_bridge_getlink)(struct sk_buff *skb, 1525 u32 pid, u32 seq, 1526 struct net_device *dev, 1527 u32 filter_mask, 1528 int nlflags); 1529 int (*ndo_bridge_dellink)(struct net_device *dev, 1530 struct nlmsghdr *nlh, 1531 u16 flags); 1532 int (*ndo_change_carrier)(struct net_device *dev, 1533 bool new_carrier); 1534 int (*ndo_get_phys_port_id)(struct net_device *dev, 1535 struct netdev_phys_item_id *ppid); 1536 int (*ndo_get_port_parent_id)(struct net_device *dev, 1537 struct netdev_phys_item_id *ppid); 1538 int (*ndo_get_phys_port_name)(struct net_device *dev, 1539 char *name, size_t len); 1540 void* (*ndo_dfwd_add_station)(struct net_device *pdev, 1541 struct net_device *dev); 1542 void (*ndo_dfwd_del_station)(struct net_device *pdev, 1543 void *priv); 1544 1545 int (*ndo_set_tx_maxrate)(struct net_device *dev, 1546 int queue_index, 1547 u32 maxrate); 1548 int (*ndo_get_iflink)(const struct net_device *dev); 1549 int (*ndo_change_proto_down)(struct net_device *dev, 1550 bool proto_down); 1551 int (*ndo_fill_metadata_dst)(struct net_device *dev, 1552 struct sk_buff *skb); 1553 void (*ndo_set_rx_headroom)(struct net_device *dev, 1554 int needed_headroom); 1555 int (*ndo_bpf)(struct net_device *dev, 1556 struct netdev_bpf *bpf); 1557 int (*ndo_xdp_xmit)(struct net_device *dev, int n, 1558 struct xdp_frame **xdp, 1559 u32 flags); 1560 int (*ndo_xsk_wakeup)(struct net_device *dev, 1561 u32 queue_id, u32 flags); 1562 struct devlink_port * (*ndo_get_devlink_port)(struct net_device *dev); 1563 int (*ndo_tunnel_ctl)(struct net_device *dev, 1564 struct ip_tunnel_parm *p, int cmd); 1565 struct net_device * (*ndo_get_peer_dev)(struct net_device *dev); 1566 int (*ndo_fill_forward_path)(struct net_device_path_ctx *ctx, 1567 struct net_device_path *path); 1568 }; 1569 1570 /** 1571 * enum netdev_priv_flags - &struct net_device priv_flags 1572 * 1573 * These are the &struct net_device, they are only set internally 1574 * by drivers and used in the kernel. These flags are invisible to 1575 * userspace; this means that the order of these flags can change 1576 * during any kernel release. 1577 * 1578 * You should have a pretty good reason to be extending these flags. 1579 * 1580 * @IFF_802_1Q_VLAN: 802.1Q VLAN device 1581 * @IFF_EBRIDGE: Ethernet bridging device 1582 * @IFF_BONDING: bonding master or slave 1583 * @IFF_ISATAP: ISATAP interface (RFC4214) 1584 * @IFF_WAN_HDLC: WAN HDLC device 1585 * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to 1586 * release skb->dst 1587 * @IFF_DONT_BRIDGE: disallow bridging this ether dev 1588 * @IFF_DISABLE_NETPOLL: disable netpoll at run-time 1589 * @IFF_MACVLAN_PORT: device used as macvlan port 1590 * @IFF_BRIDGE_PORT: device used as bridge port 1591 * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port 1592 * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit 1593 * @IFF_UNICAST_FLT: Supports unicast filtering 1594 * @IFF_TEAM_PORT: device used as team port 1595 * @IFF_SUPP_NOFCS: device supports sending custom FCS 1596 * @IFF_LIVE_ADDR_CHANGE: device supports hardware address 1597 * change when it's running 1598 * @IFF_MACVLAN: Macvlan device 1599 * @IFF_XMIT_DST_RELEASE_PERM: IFF_XMIT_DST_RELEASE not taking into account 1600 * underlying stacked devices 1601 * @IFF_L3MDEV_MASTER: device is an L3 master device 1602 * @IFF_NO_QUEUE: device can run without qdisc attached 1603 * @IFF_OPENVSWITCH: device is a Open vSwitch master 1604 * @IFF_L3MDEV_SLAVE: device is enslaved to an L3 master device 1605 * @IFF_TEAM: device is a team device 1606 * @IFF_RXFH_CONFIGURED: device has had Rx Flow indirection table configured 1607 * @IFF_PHONY_HEADROOM: the headroom value is controlled by an external 1608 * entity (i.e. the master device for bridged veth) 1609 * @IFF_MACSEC: device is a MACsec device 1610 * @IFF_NO_RX_HANDLER: device doesn't support the rx_handler hook 1611 * @IFF_FAILOVER: device is a failover master device 1612 * @IFF_FAILOVER_SLAVE: device is lower dev of a failover master device 1613 * @IFF_L3MDEV_RX_HANDLER: only invoke the rx handler of L3 master device 1614 * @IFF_LIVE_RENAME_OK: rename is allowed while device is up and running 1615 * @IFF_TX_SKB_NO_LINEAR: device/driver is capable of xmitting frames with 1616 * skb_headlen(skb) == 0 (data starts from frag0) 1617 */ 1618 enum netdev_priv_flags { 1619 IFF_802_1Q_VLAN = 1<<0, 1620 IFF_EBRIDGE = 1<<1, 1621 IFF_BONDING = 1<<2, 1622 IFF_ISATAP = 1<<3, 1623 IFF_WAN_HDLC = 1<<4, 1624 IFF_XMIT_DST_RELEASE = 1<<5, 1625 IFF_DONT_BRIDGE = 1<<6, 1626 IFF_DISABLE_NETPOLL = 1<<7, 1627 IFF_MACVLAN_PORT = 1<<8, 1628 IFF_BRIDGE_PORT = 1<<9, 1629 IFF_OVS_DATAPATH = 1<<10, 1630 IFF_TX_SKB_SHARING = 1<<11, 1631 IFF_UNICAST_FLT = 1<<12, 1632 IFF_TEAM_PORT = 1<<13, 1633 IFF_SUPP_NOFCS = 1<<14, 1634 IFF_LIVE_ADDR_CHANGE = 1<<15, 1635 IFF_MACVLAN = 1<<16, 1636 IFF_XMIT_DST_RELEASE_PERM = 1<<17, 1637 IFF_L3MDEV_MASTER = 1<<18, 1638 IFF_NO_QUEUE = 1<<19, 1639 IFF_OPENVSWITCH = 1<<20, 1640 IFF_L3MDEV_SLAVE = 1<<21, 1641 IFF_TEAM = 1<<22, 1642 IFF_RXFH_CONFIGURED = 1<<23, 1643 IFF_PHONY_HEADROOM = 1<<24, 1644 IFF_MACSEC = 1<<25, 1645 IFF_NO_RX_HANDLER = 1<<26, 1646 IFF_FAILOVER = 1<<27, 1647 IFF_FAILOVER_SLAVE = 1<<28, 1648 IFF_L3MDEV_RX_HANDLER = 1<<29, 1649 IFF_LIVE_RENAME_OK = 1<<30, 1650 IFF_TX_SKB_NO_LINEAR = 1<<31, 1651 }; 1652 1653 #define IFF_802_1Q_VLAN IFF_802_1Q_VLAN 1654 #define IFF_EBRIDGE IFF_EBRIDGE 1655 #define IFF_BONDING IFF_BONDING 1656 #define IFF_ISATAP IFF_ISATAP 1657 #define IFF_WAN_HDLC IFF_WAN_HDLC 1658 #define IFF_XMIT_DST_RELEASE IFF_XMIT_DST_RELEASE 1659 #define IFF_DONT_BRIDGE IFF_DONT_BRIDGE 1660 #define IFF_DISABLE_NETPOLL IFF_DISABLE_NETPOLL 1661 #define IFF_MACVLAN_PORT IFF_MACVLAN_PORT 1662 #define IFF_BRIDGE_PORT IFF_BRIDGE_PORT 1663 #define IFF_OVS_DATAPATH IFF_OVS_DATAPATH 1664 #define IFF_TX_SKB_SHARING IFF_TX_SKB_SHARING 1665 #define IFF_UNICAST_FLT IFF_UNICAST_FLT 1666 #define IFF_TEAM_PORT IFF_TEAM_PORT 1667 #define IFF_SUPP_NOFCS IFF_SUPP_NOFCS 1668 #define IFF_LIVE_ADDR_CHANGE IFF_LIVE_ADDR_CHANGE 1669 #define IFF_MACVLAN IFF_MACVLAN 1670 #define IFF_XMIT_DST_RELEASE_PERM IFF_XMIT_DST_RELEASE_PERM 1671 #define IFF_L3MDEV_MASTER IFF_L3MDEV_MASTER 1672 #define IFF_NO_QUEUE IFF_NO_QUEUE 1673 #define IFF_OPENVSWITCH IFF_OPENVSWITCH 1674 #define IFF_L3MDEV_SLAVE IFF_L3MDEV_SLAVE 1675 #define IFF_TEAM IFF_TEAM 1676 #define IFF_RXFH_CONFIGURED IFF_RXFH_CONFIGURED 1677 #define IFF_PHONY_HEADROOM IFF_PHONY_HEADROOM 1678 #define IFF_MACSEC IFF_MACSEC 1679 #define IFF_NO_RX_HANDLER IFF_NO_RX_HANDLER 1680 #define IFF_FAILOVER IFF_FAILOVER 1681 #define IFF_FAILOVER_SLAVE IFF_FAILOVER_SLAVE 1682 #define IFF_L3MDEV_RX_HANDLER IFF_L3MDEV_RX_HANDLER 1683 #define IFF_LIVE_RENAME_OK IFF_LIVE_RENAME_OK 1684 #define IFF_TX_SKB_NO_LINEAR IFF_TX_SKB_NO_LINEAR 1685 1686 /* Specifies the type of the struct net_device::ml_priv pointer */ 1687 enum netdev_ml_priv_type { 1688 ML_PRIV_NONE, 1689 ML_PRIV_CAN, 1690 }; 1691 1692 /** 1693 * struct net_device - The DEVICE structure. 1694 * 1695 * Actually, this whole structure is a big mistake. It mixes I/O 1696 * data with strictly "high-level" data, and it has to know about 1697 * almost every data structure used in the INET module. 1698 * 1699 * @name: This is the first field of the "visible" part of this structure 1700 * (i.e. as seen by users in the "Space.c" file). It is the name 1701 * of the interface. 1702 * 1703 * @name_node: Name hashlist node 1704 * @ifalias: SNMP alias 1705 * @mem_end: Shared memory end 1706 * @mem_start: Shared memory start 1707 * @base_addr: Device I/O address 1708 * @irq: Device IRQ number 1709 * 1710 * @state: Generic network queuing layer state, see netdev_state_t 1711 * @dev_list: The global list of network devices 1712 * @napi_list: List entry used for polling NAPI devices 1713 * @unreg_list: List entry when we are unregistering the 1714 * device; see the function unregister_netdev 1715 * @close_list: List entry used when we are closing the device 1716 * @ptype_all: Device-specific packet handlers for all protocols 1717 * @ptype_specific: Device-specific, protocol-specific packet handlers 1718 * 1719 * @adj_list: Directly linked devices, like slaves for bonding 1720 * @features: Currently active device features 1721 * @hw_features: User-changeable features 1722 * 1723 * @wanted_features: User-requested features 1724 * @vlan_features: Mask of features inheritable by VLAN devices 1725 * 1726 * @hw_enc_features: Mask of features inherited by encapsulating devices 1727 * This field indicates what encapsulation 1728 * offloads the hardware is capable of doing, 1729 * and drivers will need to set them appropriately. 1730 * 1731 * @mpls_features: Mask of features inheritable by MPLS 1732 * @gso_partial_features: value(s) from NETIF_F_GSO\* 1733 * 1734 * @ifindex: interface index 1735 * @group: The group the device belongs to 1736 * 1737 * @stats: Statistics struct, which was left as a legacy, use 1738 * rtnl_link_stats64 instead 1739 * 1740 * @rx_dropped: Dropped packets by core network, 1741 * do not use this in drivers 1742 * @tx_dropped: Dropped packets by core network, 1743 * do not use this in drivers 1744 * @rx_nohandler: nohandler dropped packets by core network on 1745 * inactive devices, do not use this in drivers 1746 * @carrier_up_count: Number of times the carrier has been up 1747 * @carrier_down_count: Number of times the carrier has been down 1748 * 1749 * @wireless_handlers: List of functions to handle Wireless Extensions, 1750 * instead of ioctl, 1751 * see <net/iw_handler.h> for details. 1752 * @wireless_data: Instance data managed by the core of wireless extensions 1753 * 1754 * @netdev_ops: Includes several pointers to callbacks, 1755 * if one wants to override the ndo_*() functions 1756 * @ethtool_ops: Management operations 1757 * @l3mdev_ops: Layer 3 master device operations 1758 * @ndisc_ops: Includes callbacks for different IPv6 neighbour 1759 * discovery handling. Necessary for e.g. 6LoWPAN. 1760 * @xfrmdev_ops: Transformation offload operations 1761 * @tlsdev_ops: Transport Layer Security offload operations 1762 * @header_ops: Includes callbacks for creating,parsing,caching,etc 1763 * of Layer 2 headers. 1764 * 1765 * @flags: Interface flags (a la BSD) 1766 * @priv_flags: Like 'flags' but invisible to userspace, 1767 * see if.h for the definitions 1768 * @gflags: Global flags ( kept as legacy ) 1769 * @padded: How much padding added by alloc_netdev() 1770 * @operstate: RFC2863 operstate 1771 * @link_mode: Mapping policy to operstate 1772 * @if_port: Selectable AUI, TP, ... 1773 * @dma: DMA channel 1774 * @mtu: Interface MTU value 1775 * @min_mtu: Interface Minimum MTU value 1776 * @max_mtu: Interface Maximum MTU value 1777 * @type: Interface hardware type 1778 * @hard_header_len: Maximum hardware header length. 1779 * @min_header_len: Minimum hardware header length 1780 * 1781 * @needed_headroom: Extra headroom the hardware may need, but not in all 1782 * cases can this be guaranteed 1783 * @needed_tailroom: Extra tailroom the hardware may need, but not in all 1784 * cases can this be guaranteed. Some cases also use 1785 * LL_MAX_HEADER instead to allocate the skb 1786 * 1787 * interface address info: 1788 * 1789 * @perm_addr: Permanent hw address 1790 * @addr_assign_type: Hw address assignment type 1791 * @addr_len: Hardware address length 1792 * @upper_level: Maximum depth level of upper devices. 1793 * @lower_level: Maximum depth level of lower devices. 1794 * @neigh_priv_len: Used in neigh_alloc() 1795 * @dev_id: Used to differentiate devices that share 1796 * the same link layer address 1797 * @dev_port: Used to differentiate devices that share 1798 * the same function 1799 * @addr_list_lock: XXX: need comments on this one 1800 * @name_assign_type: network interface name assignment type 1801 * @uc_promisc: Counter that indicates promiscuous mode 1802 * has been enabled due to the need to listen to 1803 * additional unicast addresses in a device that 1804 * does not implement ndo_set_rx_mode() 1805 * @uc: unicast mac addresses 1806 * @mc: multicast mac addresses 1807 * @dev_addrs: list of device hw addresses 1808 * @queues_kset: Group of all Kobjects in the Tx and RX queues 1809 * @promiscuity: Number of times the NIC is told to work in 1810 * promiscuous mode; if it becomes 0 the NIC will 1811 * exit promiscuous mode 1812 * @allmulti: Counter, enables or disables allmulticast mode 1813 * 1814 * @vlan_info: VLAN info 1815 * @dsa_ptr: dsa specific data 1816 * @tipc_ptr: TIPC specific data 1817 * @atalk_ptr: AppleTalk link 1818 * @ip_ptr: IPv4 specific data 1819 * @dn_ptr: DECnet specific data 1820 * @ip6_ptr: IPv6 specific data 1821 * @ax25_ptr: AX.25 specific data 1822 * @ieee80211_ptr: IEEE 802.11 specific data, assign before registering 1823 * @ieee802154_ptr: IEEE 802.15.4 low-rate Wireless Personal Area Network 1824 * device struct 1825 * @mpls_ptr: mpls_dev struct pointer 1826 * 1827 * @dev_addr: Hw address (before bcast, 1828 * because most packets are unicast) 1829 * 1830 * @_rx: Array of RX queues 1831 * @num_rx_queues: Number of RX queues 1832 * allocated at register_netdev() time 1833 * @real_num_rx_queues: Number of RX queues currently active in device 1834 * @xdp_prog: XDP sockets filter program pointer 1835 * @gro_flush_timeout: timeout for GRO layer in NAPI 1836 * @napi_defer_hard_irqs: If not zero, provides a counter that would 1837 * allow to avoid NIC hard IRQ, on busy queues. 1838 * 1839 * @rx_handler: handler for received packets 1840 * @rx_handler_data: XXX: need comments on this one 1841 * @miniq_ingress: ingress/clsact qdisc specific data for 1842 * ingress processing 1843 * @ingress_queue: XXX: need comments on this one 1844 * @nf_hooks_ingress: netfilter hooks executed for ingress packets 1845 * @broadcast: hw bcast address 1846 * 1847 * @rx_cpu_rmap: CPU reverse-mapping for RX completion interrupts, 1848 * indexed by RX queue number. Assigned by driver. 1849 * This must only be set if the ndo_rx_flow_steer 1850 * operation is defined 1851 * @index_hlist: Device index hash chain 1852 * 1853 * @_tx: Array of TX queues 1854 * @num_tx_queues: Number of TX queues allocated at alloc_netdev_mq() time 1855 * @real_num_tx_queues: Number of TX queues currently active in device 1856 * @qdisc: Root qdisc from userspace point of view 1857 * @tx_queue_len: Max frames per queue allowed 1858 * @tx_global_lock: XXX: need comments on this one 1859 * @xdp_bulkq: XDP device bulk queue 1860 * @xps_maps: all CPUs/RXQs maps for XPS device 1861 * 1862 * @xps_maps: XXX: need comments on this one 1863 * @miniq_egress: clsact qdisc specific data for 1864 * egress processing 1865 * @qdisc_hash: qdisc hash table 1866 * @watchdog_timeo: Represents the timeout that is used by 1867 * the watchdog (see dev_watchdog()) 1868 * @watchdog_timer: List of timers 1869 * 1870 * @proto_down_reason: reason a netdev interface is held down 1871 * @pcpu_refcnt: Number of references to this device 1872 * @dev_refcnt: Number of references to this device 1873 * @todo_list: Delayed register/unregister 1874 * @link_watch_list: XXX: need comments on this one 1875 * 1876 * @reg_state: Register/unregister state machine 1877 * @dismantle: Device is going to be freed 1878 * @rtnl_link_state: This enum represents the phases of creating 1879 * a new link 1880 * 1881 * @needs_free_netdev: Should unregister perform free_netdev? 1882 * @priv_destructor: Called from unregister 1883 * @npinfo: XXX: need comments on this one 1884 * @nd_net: Network namespace this network device is inside 1885 * 1886 * @ml_priv: Mid-layer private 1887 * @ml_priv_type: Mid-layer private type 1888 * @lstats: Loopback statistics 1889 * @tstats: Tunnel statistics 1890 * @dstats: Dummy statistics 1891 * @vstats: Virtual ethernet statistics 1892 * 1893 * @garp_port: GARP 1894 * @mrp_port: MRP 1895 * 1896 * @dev: Class/net/name entry 1897 * @sysfs_groups: Space for optional device, statistics and wireless 1898 * sysfs groups 1899 * 1900 * @sysfs_rx_queue_group: Space for optional per-rx queue attributes 1901 * @rtnl_link_ops: Rtnl_link_ops 1902 * 1903 * @gso_max_size: Maximum size of generic segmentation offload 1904 * @gso_max_segs: Maximum number of segments that can be passed to the 1905 * NIC for GSO 1906 * 1907 * @dcbnl_ops: Data Center Bridging netlink ops 1908 * @num_tc: Number of traffic classes in the net device 1909 * @tc_to_txq: XXX: need comments on this one 1910 * @prio_tc_map: XXX: need comments on this one 1911 * 1912 * @fcoe_ddp_xid: Max exchange id for FCoE LRO by ddp 1913 * 1914 * @priomap: XXX: need comments on this one 1915 * @phydev: Physical device may attach itself 1916 * for hardware timestamping 1917 * @sfp_bus: attached &struct sfp_bus structure. 1918 * 1919 * @qdisc_tx_busylock: lockdep class annotating Qdisc->busylock spinlock 1920 * @qdisc_running_key: lockdep class annotating Qdisc->running seqcount 1921 * 1922 * @proto_down: protocol port state information can be sent to the 1923 * switch driver and used to set the phys state of the 1924 * switch port. 1925 * 1926 * @wol_enabled: Wake-on-LAN is enabled 1927 * 1928 * @threaded: napi threaded mode is enabled 1929 * 1930 * @net_notifier_list: List of per-net netdev notifier block 1931 * that follow this device when it is moved 1932 * to another network namespace. 1933 * 1934 * @macsec_ops: MACsec offloading ops 1935 * 1936 * @udp_tunnel_nic_info: static structure describing the UDP tunnel 1937 * offload capabilities of the device 1938 * @udp_tunnel_nic: UDP tunnel offload state 1939 * @xdp_state: stores info on attached XDP BPF programs 1940 * 1941 * @nested_level: Used as as a parameter of spin_lock_nested() of 1942 * dev->addr_list_lock. 1943 * @unlink_list: As netif_addr_lock() can be called recursively, 1944 * keep a list of interfaces to be deleted. 1945 * 1946 * FIXME: cleanup struct net_device such that network protocol info 1947 * moves out. 1948 */ 1949 1950 struct net_device { 1951 char name[IFNAMSIZ]; 1952 struct netdev_name_node *name_node; 1953 struct dev_ifalias __rcu *ifalias; 1954 /* 1955 * I/O specific fields 1956 * FIXME: Merge these and struct ifmap into one 1957 */ 1958 unsigned long mem_end; 1959 unsigned long mem_start; 1960 unsigned long base_addr; 1961 1962 /* 1963 * Some hardware also needs these fields (state,dev_list, 1964 * napi_list,unreg_list,close_list) but they are not 1965 * part of the usual set specified in Space.c. 1966 */ 1967 1968 unsigned long state; 1969 1970 struct list_head dev_list; 1971 struct list_head napi_list; 1972 struct list_head unreg_list; 1973 struct list_head close_list; 1974 struct list_head ptype_all; 1975 struct list_head ptype_specific; 1976 1977 struct { 1978 struct list_head upper; 1979 struct list_head lower; 1980 } adj_list; 1981 1982 /* Read-mostly cache-line for fast-path access */ 1983 unsigned int flags; 1984 unsigned int priv_flags; 1985 const struct net_device_ops *netdev_ops; 1986 int ifindex; 1987 unsigned short gflags; 1988 unsigned short hard_header_len; 1989 1990 /* Note : dev->mtu is often read without holding a lock. 1991 * Writers usually hold RTNL. 1992 * It is recommended to use READ_ONCE() to annotate the reads, 1993 * and to use WRITE_ONCE() to annotate the writes. 1994 */ 1995 unsigned int mtu; 1996 unsigned short needed_headroom; 1997 unsigned short needed_tailroom; 1998 1999 netdev_features_t features; 2000 netdev_features_t hw_features; 2001 netdev_features_t wanted_features; 2002 netdev_features_t vlan_features; 2003 netdev_features_t hw_enc_features; 2004 netdev_features_t mpls_features; 2005 netdev_features_t gso_partial_features; 2006 2007 unsigned int min_mtu; 2008 unsigned int max_mtu; 2009 unsigned short type; 2010 unsigned char min_header_len; 2011 unsigned char name_assign_type; 2012 2013 int group; 2014 2015 struct net_device_stats stats; /* not used by modern drivers */ 2016 2017 atomic_long_t rx_dropped; 2018 atomic_long_t tx_dropped; 2019 atomic_long_t rx_nohandler; 2020 2021 /* Stats to monitor link on/off, flapping */ 2022 atomic_t carrier_up_count; 2023 atomic_t carrier_down_count; 2024 2025 #ifdef CONFIG_WIRELESS_EXT 2026 const struct iw_handler_def *wireless_handlers; 2027 struct iw_public_data *wireless_data; 2028 #endif 2029 const struct ethtool_ops *ethtool_ops; 2030 #ifdef CONFIG_NET_L3_MASTER_DEV 2031 const struct l3mdev_ops *l3mdev_ops; 2032 #endif 2033 #if IS_ENABLED(CONFIG_IPV6) 2034 const struct ndisc_ops *ndisc_ops; 2035 #endif 2036 2037 #ifdef CONFIG_XFRM_OFFLOAD 2038 const struct xfrmdev_ops *xfrmdev_ops; 2039 #endif 2040 2041 #if IS_ENABLED(CONFIG_TLS_DEVICE) 2042 const struct tlsdev_ops *tlsdev_ops; 2043 #endif 2044 2045 const struct header_ops *header_ops; 2046 2047 unsigned char operstate; 2048 unsigned char link_mode; 2049 2050 unsigned char if_port; 2051 unsigned char dma; 2052 2053 /* Interface address info. */ 2054 unsigned char perm_addr[MAX_ADDR_LEN]; 2055 unsigned char addr_assign_type; 2056 unsigned char addr_len; 2057 unsigned char upper_level; 2058 unsigned char lower_level; 2059 2060 unsigned short neigh_priv_len; 2061 unsigned short dev_id; 2062 unsigned short dev_port; 2063 unsigned short padded; 2064 2065 spinlock_t addr_list_lock; 2066 int irq; 2067 2068 struct netdev_hw_addr_list uc; 2069 struct netdev_hw_addr_list mc; 2070 struct netdev_hw_addr_list dev_addrs; 2071 2072 #ifdef CONFIG_SYSFS 2073 struct kset *queues_kset; 2074 #endif 2075 #ifdef CONFIG_LOCKDEP 2076 struct list_head unlink_list; 2077 #endif 2078 unsigned int promiscuity; 2079 unsigned int allmulti; 2080 bool uc_promisc; 2081 #ifdef CONFIG_LOCKDEP 2082 unsigned char nested_level; 2083 #endif 2084 2085 2086 /* Protocol-specific pointers */ 2087 2088 #if IS_ENABLED(CONFIG_VLAN_8021Q) 2089 struct vlan_info __rcu *vlan_info; 2090 #endif 2091 #if IS_ENABLED(CONFIG_NET_DSA) 2092 struct dsa_port *dsa_ptr; 2093 #endif 2094 #if IS_ENABLED(CONFIG_TIPC) 2095 struct tipc_bearer __rcu *tipc_ptr; 2096 #endif 2097 #if IS_ENABLED(CONFIG_IRDA) || IS_ENABLED(CONFIG_ATALK) 2098 void *atalk_ptr; 2099 #endif 2100 struct in_device __rcu *ip_ptr; 2101 #if IS_ENABLED(CONFIG_DECNET) 2102 struct dn_dev __rcu *dn_ptr; 2103 #endif 2104 struct inet6_dev __rcu *ip6_ptr; 2105 #if IS_ENABLED(CONFIG_AX25) 2106 void *ax25_ptr; 2107 #endif 2108 struct wireless_dev *ieee80211_ptr; 2109 struct wpan_dev *ieee802154_ptr; 2110 #if IS_ENABLED(CONFIG_MPLS_ROUTING) 2111 struct mpls_dev __rcu *mpls_ptr; 2112 #endif 2113 2114 /* 2115 * Cache lines mostly used on receive path (including eth_type_trans()) 2116 */ 2117 /* Interface address info used in eth_type_trans() */ 2118 unsigned char *dev_addr; 2119 2120 struct netdev_rx_queue *_rx; 2121 unsigned int num_rx_queues; 2122 unsigned int real_num_rx_queues; 2123 2124 struct bpf_prog __rcu *xdp_prog; 2125 unsigned long gro_flush_timeout; 2126 int napi_defer_hard_irqs; 2127 rx_handler_func_t __rcu *rx_handler; 2128 void __rcu *rx_handler_data; 2129 2130 #ifdef CONFIG_NET_CLS_ACT 2131 struct mini_Qdisc __rcu *miniq_ingress; 2132 #endif 2133 struct netdev_queue __rcu *ingress_queue; 2134 #ifdef CONFIG_NETFILTER_INGRESS 2135 struct nf_hook_entries __rcu *nf_hooks_ingress; 2136 #endif 2137 2138 unsigned char broadcast[MAX_ADDR_LEN]; 2139 #ifdef CONFIG_RFS_ACCEL 2140 struct cpu_rmap *rx_cpu_rmap; 2141 #endif 2142 struct hlist_node index_hlist; 2143 2144 /* 2145 * Cache lines mostly used on transmit path 2146 */ 2147 struct netdev_queue *_tx ____cacheline_aligned_in_smp; 2148 unsigned int num_tx_queues; 2149 unsigned int real_num_tx_queues; 2150 struct Qdisc *qdisc; 2151 unsigned int tx_queue_len; 2152 spinlock_t tx_global_lock; 2153 2154 struct xdp_dev_bulk_queue __percpu *xdp_bulkq; 2155 2156 #ifdef CONFIG_XPS 2157 struct xps_dev_maps __rcu *xps_maps[XPS_MAPS_MAX]; 2158 #endif 2159 #ifdef CONFIG_NET_CLS_ACT 2160 struct mini_Qdisc __rcu *miniq_egress; 2161 #endif 2162 2163 #ifdef CONFIG_NET_SCHED 2164 DECLARE_HASHTABLE (qdisc_hash, 4); 2165 #endif 2166 /* These may be needed for future network-power-down code. */ 2167 struct timer_list watchdog_timer; 2168 int watchdog_timeo; 2169 2170 u32 proto_down_reason; 2171 2172 struct list_head todo_list; 2173 2174 #ifdef CONFIG_PCPU_DEV_REFCNT 2175 int __percpu *pcpu_refcnt; 2176 #else 2177 refcount_t dev_refcnt; 2178 #endif 2179 2180 struct list_head link_watch_list; 2181 2182 enum { NETREG_UNINITIALIZED=0, 2183 NETREG_REGISTERED, /* completed register_netdevice */ 2184 NETREG_UNREGISTERING, /* called unregister_netdevice */ 2185 NETREG_UNREGISTERED, /* completed unregister todo */ 2186 NETREG_RELEASED, /* called free_netdev */ 2187 NETREG_DUMMY, /* dummy device for NAPI poll */ 2188 } reg_state:8; 2189 2190 bool dismantle; 2191 2192 enum { 2193 RTNL_LINK_INITIALIZED, 2194 RTNL_LINK_INITIALIZING, 2195 } rtnl_link_state:16; 2196 2197 bool needs_free_netdev; 2198 void (*priv_destructor)(struct net_device *dev); 2199 2200 #ifdef CONFIG_NETPOLL 2201 struct netpoll_info __rcu *npinfo; 2202 #endif 2203 2204 possible_net_t nd_net; 2205 2206 /* mid-layer private */ 2207 void *ml_priv; 2208 enum netdev_ml_priv_type ml_priv_type; 2209 2210 union { 2211 struct pcpu_lstats __percpu *lstats; 2212 struct pcpu_sw_netstats __percpu *tstats; 2213 struct pcpu_dstats __percpu *dstats; 2214 }; 2215 2216 #if IS_ENABLED(CONFIG_GARP) 2217 struct garp_port __rcu *garp_port; 2218 #endif 2219 #if IS_ENABLED(CONFIG_MRP) 2220 struct mrp_port __rcu *mrp_port; 2221 #endif 2222 2223 struct device dev; 2224 const struct attribute_group *sysfs_groups[4]; 2225 const struct attribute_group *sysfs_rx_queue_group; 2226 2227 const struct rtnl_link_ops *rtnl_link_ops; 2228 2229 /* for setting kernel sock attribute on TCP connection setup */ 2230 #define GSO_MAX_SIZE 65536 2231 unsigned int gso_max_size; 2232 #define GSO_MAX_SEGS 65535 2233 u16 gso_max_segs; 2234 2235 #ifdef CONFIG_DCB 2236 const struct dcbnl_rtnl_ops *dcbnl_ops; 2237 #endif 2238 s16 num_tc; 2239 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE]; 2240 u8 prio_tc_map[TC_BITMASK + 1]; 2241 2242 #if IS_ENABLED(CONFIG_FCOE) 2243 unsigned int fcoe_ddp_xid; 2244 #endif 2245 #if IS_ENABLED(CONFIG_CGROUP_NET_PRIO) 2246 struct netprio_map __rcu *priomap; 2247 #endif 2248 struct phy_device *phydev; 2249 struct sfp_bus *sfp_bus; 2250 struct lock_class_key *qdisc_tx_busylock; 2251 struct lock_class_key *qdisc_running_key; 2252 bool proto_down; 2253 unsigned wol_enabled:1; 2254 unsigned threaded:1; 2255 2256 struct list_head net_notifier_list; 2257 2258 #if IS_ENABLED(CONFIG_MACSEC) 2259 /* MACsec management functions */ 2260 const struct macsec_ops *macsec_ops; 2261 #endif 2262 const struct udp_tunnel_nic_info *udp_tunnel_nic_info; 2263 struct udp_tunnel_nic *udp_tunnel_nic; 2264 2265 /* protected by rtnl_lock */ 2266 struct bpf_xdp_entity xdp_state[__MAX_XDP_MODE]; 2267 }; 2268 #define to_net_dev(d) container_of(d, struct net_device, dev) 2269 2270 static inline bool netif_elide_gro(const struct net_device *dev) 2271 { 2272 if (!(dev->features & NETIF_F_GRO) || dev->xdp_prog) 2273 return true; 2274 return false; 2275 } 2276 2277 #define NETDEV_ALIGN 32 2278 2279 static inline 2280 int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio) 2281 { 2282 return dev->prio_tc_map[prio & TC_BITMASK]; 2283 } 2284 2285 static inline 2286 int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc) 2287 { 2288 if (tc >= dev->num_tc) 2289 return -EINVAL; 2290 2291 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK; 2292 return 0; 2293 } 2294 2295 int netdev_txq_to_tc(struct net_device *dev, unsigned int txq); 2296 void netdev_reset_tc(struct net_device *dev); 2297 int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset); 2298 int netdev_set_num_tc(struct net_device *dev, u8 num_tc); 2299 2300 static inline 2301 int netdev_get_num_tc(struct net_device *dev) 2302 { 2303 return dev->num_tc; 2304 } 2305 2306 static inline void net_prefetch(void *p) 2307 { 2308 prefetch(p); 2309 #if L1_CACHE_BYTES < 128 2310 prefetch((u8 *)p + L1_CACHE_BYTES); 2311 #endif 2312 } 2313 2314 static inline void net_prefetchw(void *p) 2315 { 2316 prefetchw(p); 2317 #if L1_CACHE_BYTES < 128 2318 prefetchw((u8 *)p + L1_CACHE_BYTES); 2319 #endif 2320 } 2321 2322 void netdev_unbind_sb_channel(struct net_device *dev, 2323 struct net_device *sb_dev); 2324 int netdev_bind_sb_channel_queue(struct net_device *dev, 2325 struct net_device *sb_dev, 2326 u8 tc, u16 count, u16 offset); 2327 int netdev_set_sb_channel(struct net_device *dev, u16 channel); 2328 static inline int netdev_get_sb_channel(struct net_device *dev) 2329 { 2330 return max_t(int, -dev->num_tc, 0); 2331 } 2332 2333 static inline 2334 struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev, 2335 unsigned int index) 2336 { 2337 return &dev->_tx[index]; 2338 } 2339 2340 static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev, 2341 const struct sk_buff *skb) 2342 { 2343 return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb)); 2344 } 2345 2346 static inline void netdev_for_each_tx_queue(struct net_device *dev, 2347 void (*f)(struct net_device *, 2348 struct netdev_queue *, 2349 void *), 2350 void *arg) 2351 { 2352 unsigned int i; 2353 2354 for (i = 0; i < dev->num_tx_queues; i++) 2355 f(dev, &dev->_tx[i], arg); 2356 } 2357 2358 #define netdev_lockdep_set_classes(dev) \ 2359 { \ 2360 static struct lock_class_key qdisc_tx_busylock_key; \ 2361 static struct lock_class_key qdisc_running_key; \ 2362 static struct lock_class_key qdisc_xmit_lock_key; \ 2363 static struct lock_class_key dev_addr_list_lock_key; \ 2364 unsigned int i; \ 2365 \ 2366 (dev)->qdisc_tx_busylock = &qdisc_tx_busylock_key; \ 2367 (dev)->qdisc_running_key = &qdisc_running_key; \ 2368 lockdep_set_class(&(dev)->addr_list_lock, \ 2369 &dev_addr_list_lock_key); \ 2370 for (i = 0; i < (dev)->num_tx_queues; i++) \ 2371 lockdep_set_class(&(dev)->_tx[i]._xmit_lock, \ 2372 &qdisc_xmit_lock_key); \ 2373 } 2374 2375 u16 netdev_pick_tx(struct net_device *dev, struct sk_buff *skb, 2376 struct net_device *sb_dev); 2377 struct netdev_queue *netdev_core_pick_tx(struct net_device *dev, 2378 struct sk_buff *skb, 2379 struct net_device *sb_dev); 2380 2381 /* returns the headroom that the master device needs to take in account 2382 * when forwarding to this dev 2383 */ 2384 static inline unsigned netdev_get_fwd_headroom(struct net_device *dev) 2385 { 2386 return dev->priv_flags & IFF_PHONY_HEADROOM ? 0 : dev->needed_headroom; 2387 } 2388 2389 static inline void netdev_set_rx_headroom(struct net_device *dev, int new_hr) 2390 { 2391 if (dev->netdev_ops->ndo_set_rx_headroom) 2392 dev->netdev_ops->ndo_set_rx_headroom(dev, new_hr); 2393 } 2394 2395 /* set the device rx headroom to the dev's default */ 2396 static inline void netdev_reset_rx_headroom(struct net_device *dev) 2397 { 2398 netdev_set_rx_headroom(dev, -1); 2399 } 2400 2401 static inline void *netdev_get_ml_priv(struct net_device *dev, 2402 enum netdev_ml_priv_type type) 2403 { 2404 if (dev->ml_priv_type != type) 2405 return NULL; 2406 2407 return dev->ml_priv; 2408 } 2409 2410 static inline void netdev_set_ml_priv(struct net_device *dev, 2411 void *ml_priv, 2412 enum netdev_ml_priv_type type) 2413 { 2414 WARN(dev->ml_priv_type && dev->ml_priv_type != type, 2415 "Overwriting already set ml_priv_type (%u) with different ml_priv_type (%u)!\n", 2416 dev->ml_priv_type, type); 2417 WARN(!dev->ml_priv_type && dev->ml_priv, 2418 "Overwriting already set ml_priv and ml_priv_type is ML_PRIV_NONE!\n"); 2419 2420 dev->ml_priv = ml_priv; 2421 dev->ml_priv_type = type; 2422 } 2423 2424 /* 2425 * Net namespace inlines 2426 */ 2427 static inline 2428 struct net *dev_net(const struct net_device *dev) 2429 { 2430 return read_pnet(&dev->nd_net); 2431 } 2432 2433 static inline 2434 void dev_net_set(struct net_device *dev, struct net *net) 2435 { 2436 write_pnet(&dev->nd_net, net); 2437 } 2438 2439 /** 2440 * netdev_priv - access network device private data 2441 * @dev: network device 2442 * 2443 * Get network device private data 2444 */ 2445 static inline void *netdev_priv(const struct net_device *dev) 2446 { 2447 return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN); 2448 } 2449 2450 /* Set the sysfs physical device reference for the network logical device 2451 * if set prior to registration will cause a symlink during initialization. 2452 */ 2453 #define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev)) 2454 2455 /* Set the sysfs device type for the network logical device to allow 2456 * fine-grained identification of different network device types. For 2457 * example Ethernet, Wireless LAN, Bluetooth, WiMAX etc. 2458 */ 2459 #define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype)) 2460 2461 /* Default NAPI poll() weight 2462 * Device drivers are strongly advised to not use bigger value 2463 */ 2464 #define NAPI_POLL_WEIGHT 64 2465 2466 /** 2467 * netif_napi_add - initialize a NAPI context 2468 * @dev: network device 2469 * @napi: NAPI context 2470 * @poll: polling function 2471 * @weight: default weight 2472 * 2473 * netif_napi_add() must be used to initialize a NAPI context prior to calling 2474 * *any* of the other NAPI-related functions. 2475 */ 2476 void netif_napi_add(struct net_device *dev, struct napi_struct *napi, 2477 int (*poll)(struct napi_struct *, int), int weight); 2478 2479 /** 2480 * netif_tx_napi_add - initialize a NAPI context 2481 * @dev: network device 2482 * @napi: NAPI context 2483 * @poll: polling function 2484 * @weight: default weight 2485 * 2486 * This variant of netif_napi_add() should be used from drivers using NAPI 2487 * to exclusively poll a TX queue. 2488 * This will avoid we add it into napi_hash[], thus polluting this hash table. 2489 */ 2490 static inline void netif_tx_napi_add(struct net_device *dev, 2491 struct napi_struct *napi, 2492 int (*poll)(struct napi_struct *, int), 2493 int weight) 2494 { 2495 set_bit(NAPI_STATE_NO_BUSY_POLL, &napi->state); 2496 netif_napi_add(dev, napi, poll, weight); 2497 } 2498 2499 /** 2500 * __netif_napi_del - remove a NAPI context 2501 * @napi: NAPI context 2502 * 2503 * Warning: caller must observe RCU grace period before freeing memory 2504 * containing @napi. Drivers might want to call this helper to combine 2505 * all the needed RCU grace periods into a single one. 2506 */ 2507 void __netif_napi_del(struct napi_struct *napi); 2508 2509 /** 2510 * netif_napi_del - remove a NAPI context 2511 * @napi: NAPI context 2512 * 2513 * netif_napi_del() removes a NAPI context from the network device NAPI list 2514 */ 2515 static inline void netif_napi_del(struct napi_struct *napi) 2516 { 2517 __netif_napi_del(napi); 2518 synchronize_net(); 2519 } 2520 2521 struct napi_gro_cb { 2522 /* Virtual address of skb_shinfo(skb)->frags[0].page + offset. */ 2523 void *frag0; 2524 2525 /* Length of frag0. */ 2526 unsigned int frag0_len; 2527 2528 /* This indicates where we are processing relative to skb->data. */ 2529 int data_offset; 2530 2531 /* This is non-zero if the packet cannot be merged with the new skb. */ 2532 u16 flush; 2533 2534 /* Save the IP ID here and check when we get to the transport layer */ 2535 u16 flush_id; 2536 2537 /* Number of segments aggregated. */ 2538 u16 count; 2539 2540 /* Start offset for remote checksum offload */ 2541 u16 gro_remcsum_start; 2542 2543 /* jiffies when first packet was created/queued */ 2544 unsigned long age; 2545 2546 /* Used in ipv6_gro_receive() and foo-over-udp */ 2547 u16 proto; 2548 2549 /* This is non-zero if the packet may be of the same flow. */ 2550 u8 same_flow:1; 2551 2552 /* Used in tunnel GRO receive */ 2553 u8 encap_mark:1; 2554 2555 /* GRO checksum is valid */ 2556 u8 csum_valid:1; 2557 2558 /* Number of checksums via CHECKSUM_UNNECESSARY */ 2559 u8 csum_cnt:3; 2560 2561 /* Free the skb? */ 2562 u8 free:2; 2563 #define NAPI_GRO_FREE 1 2564 #define NAPI_GRO_FREE_STOLEN_HEAD 2 2565 2566 /* Used in foo-over-udp, set in udp[46]_gro_receive */ 2567 u8 is_ipv6:1; 2568 2569 /* Used in GRE, set in fou/gue_gro_receive */ 2570 u8 is_fou:1; 2571 2572 /* Used to determine if flush_id can be ignored */ 2573 u8 is_atomic:1; 2574 2575 /* Number of gro_receive callbacks this packet already went through */ 2576 u8 recursion_counter:4; 2577 2578 /* GRO is done by frag_list pointer chaining. */ 2579 u8 is_flist:1; 2580 2581 /* used to support CHECKSUM_COMPLETE for tunneling protocols */ 2582 __wsum csum; 2583 2584 /* used in skb_gro_receive() slow path */ 2585 struct sk_buff *last; 2586 }; 2587 2588 #define NAPI_GRO_CB(skb) ((struct napi_gro_cb *)(skb)->cb) 2589 2590 #define GRO_RECURSION_LIMIT 15 2591 static inline int gro_recursion_inc_test(struct sk_buff *skb) 2592 { 2593 return ++NAPI_GRO_CB(skb)->recursion_counter == GRO_RECURSION_LIMIT; 2594 } 2595 2596 typedef struct sk_buff *(*gro_receive_t)(struct list_head *, struct sk_buff *); 2597 static inline struct sk_buff *call_gro_receive(gro_receive_t cb, 2598 struct list_head *head, 2599 struct sk_buff *skb) 2600 { 2601 if (unlikely(gro_recursion_inc_test(skb))) { 2602 NAPI_GRO_CB(skb)->flush |= 1; 2603 return NULL; 2604 } 2605 2606 return cb(head, skb); 2607 } 2608 2609 typedef struct sk_buff *(*gro_receive_sk_t)(struct sock *, struct list_head *, 2610 struct sk_buff *); 2611 static inline struct sk_buff *call_gro_receive_sk(gro_receive_sk_t cb, 2612 struct sock *sk, 2613 struct list_head *head, 2614 struct sk_buff *skb) 2615 { 2616 if (unlikely(gro_recursion_inc_test(skb))) { 2617 NAPI_GRO_CB(skb)->flush |= 1; 2618 return NULL; 2619 } 2620 2621 return cb(sk, head, skb); 2622 } 2623 2624 struct packet_type { 2625 __be16 type; /* This is really htons(ether_type). */ 2626 bool ignore_outgoing; 2627 struct net_device *dev; /* NULL is wildcarded here */ 2628 int (*func) (struct sk_buff *, 2629 struct net_device *, 2630 struct packet_type *, 2631 struct net_device *); 2632 void (*list_func) (struct list_head *, 2633 struct packet_type *, 2634 struct net_device *); 2635 bool (*id_match)(struct packet_type *ptype, 2636 struct sock *sk); 2637 void *af_packet_priv; 2638 struct list_head list; 2639 }; 2640 2641 struct offload_callbacks { 2642 struct sk_buff *(*gso_segment)(struct sk_buff *skb, 2643 netdev_features_t features); 2644 struct sk_buff *(*gro_receive)(struct list_head *head, 2645 struct sk_buff *skb); 2646 int (*gro_complete)(struct sk_buff *skb, int nhoff); 2647 }; 2648 2649 struct packet_offload { 2650 __be16 type; /* This is really htons(ether_type). */ 2651 u16 priority; 2652 struct offload_callbacks callbacks; 2653 struct list_head list; 2654 }; 2655 2656 /* often modified stats are per-CPU, other are shared (netdev->stats) */ 2657 struct pcpu_sw_netstats { 2658 u64 rx_packets; 2659 u64 rx_bytes; 2660 u64 tx_packets; 2661 u64 tx_bytes; 2662 struct u64_stats_sync syncp; 2663 } __aligned(4 * sizeof(u64)); 2664 2665 struct pcpu_lstats { 2666 u64_stats_t packets; 2667 u64_stats_t bytes; 2668 struct u64_stats_sync syncp; 2669 } __aligned(2 * sizeof(u64)); 2670 2671 void dev_lstats_read(struct net_device *dev, u64 *packets, u64 *bytes); 2672 2673 static inline void dev_sw_netstats_rx_add(struct net_device *dev, unsigned int len) 2674 { 2675 struct pcpu_sw_netstats *tstats = this_cpu_ptr(dev->tstats); 2676 2677 u64_stats_update_begin(&tstats->syncp); 2678 tstats->rx_bytes += len; 2679 tstats->rx_packets++; 2680 u64_stats_update_end(&tstats->syncp); 2681 } 2682 2683 static inline void dev_sw_netstats_tx_add(struct net_device *dev, 2684 unsigned int packets, 2685 unsigned int len) 2686 { 2687 struct pcpu_sw_netstats *tstats = this_cpu_ptr(dev->tstats); 2688 2689 u64_stats_update_begin(&tstats->syncp); 2690 tstats->tx_bytes += len; 2691 tstats->tx_packets += packets; 2692 u64_stats_update_end(&tstats->syncp); 2693 } 2694 2695 static inline void dev_lstats_add(struct net_device *dev, unsigned int len) 2696 { 2697 struct pcpu_lstats *lstats = this_cpu_ptr(dev->lstats); 2698 2699 u64_stats_update_begin(&lstats->syncp); 2700 u64_stats_add(&lstats->bytes, len); 2701 u64_stats_inc(&lstats->packets); 2702 u64_stats_update_end(&lstats->syncp); 2703 } 2704 2705 #define __netdev_alloc_pcpu_stats(type, gfp) \ 2706 ({ \ 2707 typeof(type) __percpu *pcpu_stats = alloc_percpu_gfp(type, gfp);\ 2708 if (pcpu_stats) { \ 2709 int __cpu; \ 2710 for_each_possible_cpu(__cpu) { \ 2711 typeof(type) *stat; \ 2712 stat = per_cpu_ptr(pcpu_stats, __cpu); \ 2713 u64_stats_init(&stat->syncp); \ 2714 } \ 2715 } \ 2716 pcpu_stats; \ 2717 }) 2718 2719 #define netdev_alloc_pcpu_stats(type) \ 2720 __netdev_alloc_pcpu_stats(type, GFP_KERNEL) 2721 2722 #define devm_netdev_alloc_pcpu_stats(dev, type) \ 2723 ({ \ 2724 typeof(type) __percpu *pcpu_stats = devm_alloc_percpu(dev, type);\ 2725 if (pcpu_stats) { \ 2726 int __cpu; \ 2727 for_each_possible_cpu(__cpu) { \ 2728 typeof(type) *stat; \ 2729 stat = per_cpu_ptr(pcpu_stats, __cpu); \ 2730 u64_stats_init(&stat->syncp); \ 2731 } \ 2732 } \ 2733 pcpu_stats; \ 2734 }) 2735 2736 enum netdev_lag_tx_type { 2737 NETDEV_LAG_TX_TYPE_UNKNOWN, 2738 NETDEV_LAG_TX_TYPE_RANDOM, 2739 NETDEV_LAG_TX_TYPE_BROADCAST, 2740 NETDEV_LAG_TX_TYPE_ROUNDROBIN, 2741 NETDEV_LAG_TX_TYPE_ACTIVEBACKUP, 2742 NETDEV_LAG_TX_TYPE_HASH, 2743 }; 2744 2745 enum netdev_lag_hash { 2746 NETDEV_LAG_HASH_NONE, 2747 NETDEV_LAG_HASH_L2, 2748 NETDEV_LAG_HASH_L34, 2749 NETDEV_LAG_HASH_L23, 2750 NETDEV_LAG_HASH_E23, 2751 NETDEV_LAG_HASH_E34, 2752 NETDEV_LAG_HASH_VLAN_SRCMAC, 2753 NETDEV_LAG_HASH_UNKNOWN, 2754 }; 2755 2756 struct netdev_lag_upper_info { 2757 enum netdev_lag_tx_type tx_type; 2758 enum netdev_lag_hash hash_type; 2759 }; 2760 2761 struct netdev_lag_lower_state_info { 2762 u8 link_up : 1, 2763 tx_enabled : 1; 2764 }; 2765 2766 #include <linux/notifier.h> 2767 2768 /* netdevice notifier chain. Please remember to update netdev_cmd_to_name() 2769 * and the rtnetlink notification exclusion list in rtnetlink_event() when 2770 * adding new types. 2771 */ 2772 enum netdev_cmd { 2773 NETDEV_UP = 1, /* For now you can't veto a device up/down */ 2774 NETDEV_DOWN, 2775 NETDEV_REBOOT, /* Tell a protocol stack a network interface 2776 detected a hardware crash and restarted 2777 - we can use this eg to kick tcp sessions 2778 once done */ 2779 NETDEV_CHANGE, /* Notify device state change */ 2780 NETDEV_REGISTER, 2781 NETDEV_UNREGISTER, 2782 NETDEV_CHANGEMTU, /* notify after mtu change happened */ 2783 NETDEV_CHANGEADDR, /* notify after the address change */ 2784 NETDEV_PRE_CHANGEADDR, /* notify before the address change */ 2785 NETDEV_GOING_DOWN, 2786 NETDEV_CHANGENAME, 2787 NETDEV_FEAT_CHANGE, 2788 NETDEV_BONDING_FAILOVER, 2789 NETDEV_PRE_UP, 2790 NETDEV_PRE_TYPE_CHANGE, 2791 NETDEV_POST_TYPE_CHANGE, 2792 NETDEV_POST_INIT, 2793 NETDEV_RELEASE, 2794 NETDEV_NOTIFY_PEERS, 2795 NETDEV_JOIN, 2796 NETDEV_CHANGEUPPER, 2797 NETDEV_RESEND_IGMP, 2798 NETDEV_PRECHANGEMTU, /* notify before mtu change happened */ 2799 NETDEV_CHANGEINFODATA, 2800 NETDEV_BONDING_INFO, 2801 NETDEV_PRECHANGEUPPER, 2802 NETDEV_CHANGELOWERSTATE, 2803 NETDEV_UDP_TUNNEL_PUSH_INFO, 2804 NETDEV_UDP_TUNNEL_DROP_INFO, 2805 NETDEV_CHANGE_TX_QUEUE_LEN, 2806 NETDEV_CVLAN_FILTER_PUSH_INFO, 2807 NETDEV_CVLAN_FILTER_DROP_INFO, 2808 NETDEV_SVLAN_FILTER_PUSH_INFO, 2809 NETDEV_SVLAN_FILTER_DROP_INFO, 2810 }; 2811 const char *netdev_cmd_to_name(enum netdev_cmd cmd); 2812 2813 int register_netdevice_notifier(struct notifier_block *nb); 2814 int unregister_netdevice_notifier(struct notifier_block *nb); 2815 int register_netdevice_notifier_net(struct net *net, struct notifier_block *nb); 2816 int unregister_netdevice_notifier_net(struct net *net, 2817 struct notifier_block *nb); 2818 int register_netdevice_notifier_dev_net(struct net_device *dev, 2819 struct notifier_block *nb, 2820 struct netdev_net_notifier *nn); 2821 int unregister_netdevice_notifier_dev_net(struct net_device *dev, 2822 struct notifier_block *nb, 2823 struct netdev_net_notifier *nn); 2824 2825 struct netdev_notifier_info { 2826 struct net_device *dev; 2827 struct netlink_ext_ack *extack; 2828 }; 2829 2830 struct netdev_notifier_info_ext { 2831 struct netdev_notifier_info info; /* must be first */ 2832 union { 2833 u32 mtu; 2834 } ext; 2835 }; 2836 2837 struct netdev_notifier_change_info { 2838 struct netdev_notifier_info info; /* must be first */ 2839 unsigned int flags_changed; 2840 }; 2841 2842 struct netdev_notifier_changeupper_info { 2843 struct netdev_notifier_info info; /* must be first */ 2844 struct net_device *upper_dev; /* new upper dev */ 2845 bool master; /* is upper dev master */ 2846 bool linking; /* is the notification for link or unlink */ 2847 void *upper_info; /* upper dev info */ 2848 }; 2849 2850 struct netdev_notifier_changelowerstate_info { 2851 struct netdev_notifier_info info; /* must be first */ 2852 void *lower_state_info; /* is lower dev state */ 2853 }; 2854 2855 struct netdev_notifier_pre_changeaddr_info { 2856 struct netdev_notifier_info info; /* must be first */ 2857 const unsigned char *dev_addr; 2858 }; 2859 2860 static inline void netdev_notifier_info_init(struct netdev_notifier_info *info, 2861 struct net_device *dev) 2862 { 2863 info->dev = dev; 2864 info->extack = NULL; 2865 } 2866 2867 static inline struct net_device * 2868 netdev_notifier_info_to_dev(const struct netdev_notifier_info *info) 2869 { 2870 return info->dev; 2871 } 2872 2873 static inline struct netlink_ext_ack * 2874 netdev_notifier_info_to_extack(const struct netdev_notifier_info *info) 2875 { 2876 return info->extack; 2877 } 2878 2879 int call_netdevice_notifiers(unsigned long val, struct net_device *dev); 2880 2881 2882 extern rwlock_t dev_base_lock; /* Device list lock */ 2883 2884 #define for_each_netdev(net, d) \ 2885 list_for_each_entry(d, &(net)->dev_base_head, dev_list) 2886 #define for_each_netdev_reverse(net, d) \ 2887 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list) 2888 #define for_each_netdev_rcu(net, d) \ 2889 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list) 2890 #define for_each_netdev_safe(net, d, n) \ 2891 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list) 2892 #define for_each_netdev_continue(net, d) \ 2893 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list) 2894 #define for_each_netdev_continue_reverse(net, d) \ 2895 list_for_each_entry_continue_reverse(d, &(net)->dev_base_head, \ 2896 dev_list) 2897 #define for_each_netdev_continue_rcu(net, d) \ 2898 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list) 2899 #define for_each_netdev_in_bond_rcu(bond, slave) \ 2900 for_each_netdev_rcu(&init_net, slave) \ 2901 if (netdev_master_upper_dev_get_rcu(slave) == (bond)) 2902 #define net_device_entry(lh) list_entry(lh, struct net_device, dev_list) 2903 2904 static inline struct net_device *next_net_device(struct net_device *dev) 2905 { 2906 struct list_head *lh; 2907 struct net *net; 2908 2909 net = dev_net(dev); 2910 lh = dev->dev_list.next; 2911 return lh == &net->dev_base_head ? NULL : net_device_entry(lh); 2912 } 2913 2914 static inline struct net_device *next_net_device_rcu(struct net_device *dev) 2915 { 2916 struct list_head *lh; 2917 struct net *net; 2918 2919 net = dev_net(dev); 2920 lh = rcu_dereference(list_next_rcu(&dev->dev_list)); 2921 return lh == &net->dev_base_head ? NULL : net_device_entry(lh); 2922 } 2923 2924 static inline struct net_device *first_net_device(struct net *net) 2925 { 2926 return list_empty(&net->dev_base_head) ? NULL : 2927 net_device_entry(net->dev_base_head.next); 2928 } 2929 2930 static inline struct net_device *first_net_device_rcu(struct net *net) 2931 { 2932 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head)); 2933 2934 return lh == &net->dev_base_head ? NULL : net_device_entry(lh); 2935 } 2936 2937 int netdev_boot_setup_check(struct net_device *dev); 2938 unsigned long netdev_boot_base(const char *prefix, int unit); 2939 struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type, 2940 const char *hwaddr); 2941 struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type); 2942 void dev_add_pack(struct packet_type *pt); 2943 void dev_remove_pack(struct packet_type *pt); 2944 void __dev_remove_pack(struct packet_type *pt); 2945 void dev_add_offload(struct packet_offload *po); 2946 void dev_remove_offload(struct packet_offload *po); 2947 2948 int dev_get_iflink(const struct net_device *dev); 2949 int dev_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb); 2950 int dev_fill_forward_path(const struct net_device *dev, const u8 *daddr, 2951 struct net_device_path_stack *stack); 2952 struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags, 2953 unsigned short mask); 2954 struct net_device *dev_get_by_name(struct net *net, const char *name); 2955 struct net_device *dev_get_by_name_rcu(struct net *net, const char *name); 2956 struct net_device *__dev_get_by_name(struct net *net, const char *name); 2957 int dev_alloc_name(struct net_device *dev, const char *name); 2958 int dev_open(struct net_device *dev, struct netlink_ext_ack *extack); 2959 void dev_close(struct net_device *dev); 2960 void dev_close_many(struct list_head *head, bool unlink); 2961 void dev_disable_lro(struct net_device *dev); 2962 int dev_loopback_xmit(struct net *net, struct sock *sk, struct sk_buff *newskb); 2963 u16 dev_pick_tx_zero(struct net_device *dev, struct sk_buff *skb, 2964 struct net_device *sb_dev); 2965 u16 dev_pick_tx_cpu_id(struct net_device *dev, struct sk_buff *skb, 2966 struct net_device *sb_dev); 2967 2968 int dev_queue_xmit(struct sk_buff *skb); 2969 int dev_queue_xmit_accel(struct sk_buff *skb, struct net_device *sb_dev); 2970 int __dev_direct_xmit(struct sk_buff *skb, u16 queue_id); 2971 2972 static inline int dev_direct_xmit(struct sk_buff *skb, u16 queue_id) 2973 { 2974 int ret; 2975 2976 ret = __dev_direct_xmit(skb, queue_id); 2977 if (!dev_xmit_complete(ret)) 2978 kfree_skb(skb); 2979 return ret; 2980 } 2981 2982 int register_netdevice(struct net_device *dev); 2983 void unregister_netdevice_queue(struct net_device *dev, struct list_head *head); 2984 void unregister_netdevice_many(struct list_head *head); 2985 static inline void unregister_netdevice(struct net_device *dev) 2986 { 2987 unregister_netdevice_queue(dev, NULL); 2988 } 2989 2990 int netdev_refcnt_read(const struct net_device *dev); 2991 void free_netdev(struct net_device *dev); 2992 void netdev_freemem(struct net_device *dev); 2993 int init_dummy_netdev(struct net_device *dev); 2994 2995 struct net_device *netdev_get_xmit_slave(struct net_device *dev, 2996 struct sk_buff *skb, 2997 bool all_slaves); 2998 struct net_device *netdev_sk_get_lowest_dev(struct net_device *dev, 2999 struct sock *sk); 3000 struct net_device *dev_get_by_index(struct net *net, int ifindex); 3001 struct net_device *__dev_get_by_index(struct net *net, int ifindex); 3002 struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex); 3003 struct net_device *dev_get_by_napi_id(unsigned int napi_id); 3004 int netdev_get_name(struct net *net, char *name, int ifindex); 3005 int dev_restart(struct net_device *dev); 3006 int skb_gro_receive(struct sk_buff *p, struct sk_buff *skb); 3007 int skb_gro_receive_list(struct sk_buff *p, struct sk_buff *skb); 3008 3009 static inline unsigned int skb_gro_offset(const struct sk_buff *skb) 3010 { 3011 return NAPI_GRO_CB(skb)->data_offset; 3012 } 3013 3014 static inline unsigned int skb_gro_len(const struct sk_buff *skb) 3015 { 3016 return skb->len - NAPI_GRO_CB(skb)->data_offset; 3017 } 3018 3019 static inline void skb_gro_pull(struct sk_buff *skb, unsigned int len) 3020 { 3021 NAPI_GRO_CB(skb)->data_offset += len; 3022 } 3023 3024 static inline void *skb_gro_header_fast(struct sk_buff *skb, 3025 unsigned int offset) 3026 { 3027 return NAPI_GRO_CB(skb)->frag0 + offset; 3028 } 3029 3030 static inline int skb_gro_header_hard(struct sk_buff *skb, unsigned int hlen) 3031 { 3032 return NAPI_GRO_CB(skb)->frag0_len < hlen; 3033 } 3034 3035 static inline void skb_gro_frag0_invalidate(struct sk_buff *skb) 3036 { 3037 NAPI_GRO_CB(skb)->frag0 = NULL; 3038 NAPI_GRO_CB(skb)->frag0_len = 0; 3039 } 3040 3041 static inline void *skb_gro_header_slow(struct sk_buff *skb, unsigned int hlen, 3042 unsigned int offset) 3043 { 3044 if (!pskb_may_pull(skb, hlen)) 3045 return NULL; 3046 3047 skb_gro_frag0_invalidate(skb); 3048 return skb->data + offset; 3049 } 3050 3051 static inline void *skb_gro_network_header(struct sk_buff *skb) 3052 { 3053 return (NAPI_GRO_CB(skb)->frag0 ?: skb->data) + 3054 skb_network_offset(skb); 3055 } 3056 3057 static inline void skb_gro_postpull_rcsum(struct sk_buff *skb, 3058 const void *start, unsigned int len) 3059 { 3060 if (NAPI_GRO_CB(skb)->csum_valid) 3061 NAPI_GRO_CB(skb)->csum = csum_sub(NAPI_GRO_CB(skb)->csum, 3062 csum_partial(start, len, 0)); 3063 } 3064 3065 /* GRO checksum functions. These are logical equivalents of the normal 3066 * checksum functions (in skbuff.h) except that they operate on the GRO 3067 * offsets and fields in sk_buff. 3068 */ 3069 3070 __sum16 __skb_gro_checksum_complete(struct sk_buff *skb); 3071 3072 static inline bool skb_at_gro_remcsum_start(struct sk_buff *skb) 3073 { 3074 return (NAPI_GRO_CB(skb)->gro_remcsum_start == skb_gro_offset(skb)); 3075 } 3076 3077 static inline bool __skb_gro_checksum_validate_needed(struct sk_buff *skb, 3078 bool zero_okay, 3079 __sum16 check) 3080 { 3081 return ((skb->ip_summed != CHECKSUM_PARTIAL || 3082 skb_checksum_start_offset(skb) < 3083 skb_gro_offset(skb)) && 3084 !skb_at_gro_remcsum_start(skb) && 3085 NAPI_GRO_CB(skb)->csum_cnt == 0 && 3086 (!zero_okay || check)); 3087 } 3088 3089 static inline __sum16 __skb_gro_checksum_validate_complete(struct sk_buff *skb, 3090 __wsum psum) 3091 { 3092 if (NAPI_GRO_CB(skb)->csum_valid && 3093 !csum_fold(csum_add(psum, NAPI_GRO_CB(skb)->csum))) 3094 return 0; 3095 3096 NAPI_GRO_CB(skb)->csum = psum; 3097 3098 return __skb_gro_checksum_complete(skb); 3099 } 3100 3101 static inline void skb_gro_incr_csum_unnecessary(struct sk_buff *skb) 3102 { 3103 if (NAPI_GRO_CB(skb)->csum_cnt > 0) { 3104 /* Consume a checksum from CHECKSUM_UNNECESSARY */ 3105 NAPI_GRO_CB(skb)->csum_cnt--; 3106 } else { 3107 /* Update skb for CHECKSUM_UNNECESSARY and csum_level when we 3108 * verified a new top level checksum or an encapsulated one 3109 * during GRO. This saves work if we fallback to normal path. 3110 */ 3111 __skb_incr_checksum_unnecessary(skb); 3112 } 3113 } 3114 3115 #define __skb_gro_checksum_validate(skb, proto, zero_okay, check, \ 3116 compute_pseudo) \ 3117 ({ \ 3118 __sum16 __ret = 0; \ 3119 if (__skb_gro_checksum_validate_needed(skb, zero_okay, check)) \ 3120 __ret = __skb_gro_checksum_validate_complete(skb, \ 3121 compute_pseudo(skb, proto)); \ 3122 if (!__ret) \ 3123 skb_gro_incr_csum_unnecessary(skb); \ 3124 __ret; \ 3125 }) 3126 3127 #define skb_gro_checksum_validate(skb, proto, compute_pseudo) \ 3128 __skb_gro_checksum_validate(skb, proto, false, 0, compute_pseudo) 3129 3130 #define skb_gro_checksum_validate_zero_check(skb, proto, check, \ 3131 compute_pseudo) \ 3132 __skb_gro_checksum_validate(skb, proto, true, check, compute_pseudo) 3133 3134 #define skb_gro_checksum_simple_validate(skb) \ 3135 __skb_gro_checksum_validate(skb, 0, false, 0, null_compute_pseudo) 3136 3137 static inline bool __skb_gro_checksum_convert_check(struct sk_buff *skb) 3138 { 3139 return (NAPI_GRO_CB(skb)->csum_cnt == 0 && 3140 !NAPI_GRO_CB(skb)->csum_valid); 3141 } 3142 3143 static inline void __skb_gro_checksum_convert(struct sk_buff *skb, 3144 __wsum pseudo) 3145 { 3146 NAPI_GRO_CB(skb)->csum = ~pseudo; 3147 NAPI_GRO_CB(skb)->csum_valid = 1; 3148 } 3149 3150 #define skb_gro_checksum_try_convert(skb, proto, compute_pseudo) \ 3151 do { \ 3152 if (__skb_gro_checksum_convert_check(skb)) \ 3153 __skb_gro_checksum_convert(skb, \ 3154 compute_pseudo(skb, proto)); \ 3155 } while (0) 3156 3157 struct gro_remcsum { 3158 int offset; 3159 __wsum delta; 3160 }; 3161 3162 static inline void skb_gro_remcsum_init(struct gro_remcsum *grc) 3163 { 3164 grc->offset = 0; 3165 grc->delta = 0; 3166 } 3167 3168 static inline void *skb_gro_remcsum_process(struct sk_buff *skb, void *ptr, 3169 unsigned int off, size_t hdrlen, 3170 int start, int offset, 3171 struct gro_remcsum *grc, 3172 bool nopartial) 3173 { 3174 __wsum delta; 3175 size_t plen = hdrlen + max_t(size_t, offset + sizeof(u16), start); 3176 3177 BUG_ON(!NAPI_GRO_CB(skb)->csum_valid); 3178 3179 if (!nopartial) { 3180 NAPI_GRO_CB(skb)->gro_remcsum_start = off + hdrlen + start; 3181 return ptr; 3182 } 3183 3184 ptr = skb_gro_header_fast(skb, off); 3185 if (skb_gro_header_hard(skb, off + plen)) { 3186 ptr = skb_gro_header_slow(skb, off + plen, off); 3187 if (!ptr) 3188 return NULL; 3189 } 3190 3191 delta = remcsum_adjust(ptr + hdrlen, NAPI_GRO_CB(skb)->csum, 3192 start, offset); 3193 3194 /* Adjust skb->csum since we changed the packet */ 3195 NAPI_GRO_CB(skb)->csum = csum_add(NAPI_GRO_CB(skb)->csum, delta); 3196 3197 grc->offset = off + hdrlen + offset; 3198 grc->delta = delta; 3199 3200 return ptr; 3201 } 3202 3203 static inline void skb_gro_remcsum_cleanup(struct sk_buff *skb, 3204 struct gro_remcsum *grc) 3205 { 3206 void *ptr; 3207 size_t plen = grc->offset + sizeof(u16); 3208 3209 if (!grc->delta) 3210 return; 3211 3212 ptr = skb_gro_header_fast(skb, grc->offset); 3213 if (skb_gro_header_hard(skb, grc->offset + sizeof(u16))) { 3214 ptr = skb_gro_header_slow(skb, plen, grc->offset); 3215 if (!ptr) 3216 return; 3217 } 3218 3219 remcsum_unadjust((__sum16 *)ptr, grc->delta); 3220 } 3221 3222 #ifdef CONFIG_XFRM_OFFLOAD 3223 static inline void skb_gro_flush_final(struct sk_buff *skb, struct sk_buff *pp, int flush) 3224 { 3225 if (PTR_ERR(pp) != -EINPROGRESS) 3226 NAPI_GRO_CB(skb)->flush |= flush; 3227 } 3228 static inline void skb_gro_flush_final_remcsum(struct sk_buff *skb, 3229 struct sk_buff *pp, 3230 int flush, 3231 struct gro_remcsum *grc) 3232 { 3233 if (PTR_ERR(pp) != -EINPROGRESS) { 3234 NAPI_GRO_CB(skb)->flush |= flush; 3235 skb_gro_remcsum_cleanup(skb, grc); 3236 skb->remcsum_offload = 0; 3237 } 3238 } 3239 #else 3240 static inline void skb_gro_flush_final(struct sk_buff *skb, struct sk_buff *pp, int flush) 3241 { 3242 NAPI_GRO_CB(skb)->flush |= flush; 3243 } 3244 static inline void skb_gro_flush_final_remcsum(struct sk_buff *skb, 3245 struct sk_buff *pp, 3246 int flush, 3247 struct gro_remcsum *grc) 3248 { 3249 NAPI_GRO_CB(skb)->flush |= flush; 3250 skb_gro_remcsum_cleanup(skb, grc); 3251 skb->remcsum_offload = 0; 3252 } 3253 #endif 3254 3255 static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev, 3256 unsigned short type, 3257 const void *daddr, const void *saddr, 3258 unsigned int len) 3259 { 3260 if (!dev->header_ops || !dev->header_ops->create) 3261 return 0; 3262 3263 return dev->header_ops->create(skb, dev, type, daddr, saddr, len); 3264 } 3265 3266 static inline int dev_parse_header(const struct sk_buff *skb, 3267 unsigned char *haddr) 3268 { 3269 const struct net_device *dev = skb->dev; 3270 3271 if (!dev->header_ops || !dev->header_ops->parse) 3272 return 0; 3273 return dev->header_ops->parse(skb, haddr); 3274 } 3275 3276 static inline __be16 dev_parse_header_protocol(const struct sk_buff *skb) 3277 { 3278 const struct net_device *dev = skb->dev; 3279 3280 if (!dev->header_ops || !dev->header_ops->parse_protocol) 3281 return 0; 3282 return dev->header_ops->parse_protocol(skb); 3283 } 3284 3285 /* ll_header must have at least hard_header_len allocated */ 3286 static inline bool dev_validate_header(const struct net_device *dev, 3287 char *ll_header, int len) 3288 { 3289 if (likely(len >= dev->hard_header_len)) 3290 return true; 3291 if (len < dev->min_header_len) 3292 return false; 3293 3294 if (capable(CAP_SYS_RAWIO)) { 3295 memset(ll_header + len, 0, dev->hard_header_len - len); 3296 return true; 3297 } 3298 3299 if (dev->header_ops && dev->header_ops->validate) 3300 return dev->header_ops->validate(ll_header, len); 3301 3302 return false; 3303 } 3304 3305 static inline bool dev_has_header(const struct net_device *dev) 3306 { 3307 return dev->header_ops && dev->header_ops->create; 3308 } 3309 3310 #ifdef CONFIG_NET_FLOW_LIMIT 3311 #define FLOW_LIMIT_HISTORY (1 << 7) /* must be ^2 and !overflow buckets */ 3312 struct sd_flow_limit { 3313 u64 count; 3314 unsigned int num_buckets; 3315 unsigned int history_head; 3316 u16 history[FLOW_LIMIT_HISTORY]; 3317 u8 buckets[]; 3318 }; 3319 3320 extern int netdev_flow_limit_table_len; 3321 #endif /* CONFIG_NET_FLOW_LIMIT */ 3322 3323 /* 3324 * Incoming packets are placed on per-CPU queues 3325 */ 3326 struct softnet_data { 3327 struct list_head poll_list; 3328 struct sk_buff_head process_queue; 3329 3330 /* stats */ 3331 unsigned int processed; 3332 unsigned int time_squeeze; 3333 unsigned int received_rps; 3334 #ifdef CONFIG_RPS 3335 struct softnet_data *rps_ipi_list; 3336 #endif 3337 #ifdef CONFIG_NET_FLOW_LIMIT 3338 struct sd_flow_limit __rcu *flow_limit; 3339 #endif 3340 struct Qdisc *output_queue; 3341 struct Qdisc **output_queue_tailp; 3342 struct sk_buff *completion_queue; 3343 #ifdef CONFIG_XFRM_OFFLOAD 3344 struct sk_buff_head xfrm_backlog; 3345 #endif 3346 /* written and read only by owning cpu: */ 3347 struct { 3348 u16 recursion; 3349 u8 more; 3350 } xmit; 3351 #ifdef CONFIG_RPS 3352 /* input_queue_head should be written by cpu owning this struct, 3353 * and only read by other cpus. Worth using a cache line. 3354 */ 3355 unsigned int input_queue_head ____cacheline_aligned_in_smp; 3356 3357 /* Elements below can be accessed between CPUs for RPS/RFS */ 3358 call_single_data_t csd ____cacheline_aligned_in_smp; 3359 struct softnet_data *rps_ipi_next; 3360 unsigned int cpu; 3361 unsigned int input_queue_tail; 3362 #endif 3363 unsigned int dropped; 3364 struct sk_buff_head input_pkt_queue; 3365 struct napi_struct backlog; 3366 3367 }; 3368 3369 static inline void input_queue_head_incr(struct softnet_data *sd) 3370 { 3371 #ifdef CONFIG_RPS 3372 sd->input_queue_head++; 3373 #endif 3374 } 3375 3376 static inline void input_queue_tail_incr_save(struct softnet_data *sd, 3377 unsigned int *qtail) 3378 { 3379 #ifdef CONFIG_RPS 3380 *qtail = ++sd->input_queue_tail; 3381 #endif 3382 } 3383 3384 DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data); 3385 3386 static inline int dev_recursion_level(void) 3387 { 3388 return this_cpu_read(softnet_data.xmit.recursion); 3389 } 3390 3391 #define XMIT_RECURSION_LIMIT 8 3392 static inline bool dev_xmit_recursion(void) 3393 { 3394 return unlikely(__this_cpu_read(softnet_data.xmit.recursion) > 3395 XMIT_RECURSION_LIMIT); 3396 } 3397 3398 static inline void dev_xmit_recursion_inc(void) 3399 { 3400 __this_cpu_inc(softnet_data.xmit.recursion); 3401 } 3402 3403 static inline void dev_xmit_recursion_dec(void) 3404 { 3405 __this_cpu_dec(softnet_data.xmit.recursion); 3406 } 3407 3408 void __netif_schedule(struct Qdisc *q); 3409 void netif_schedule_queue(struct netdev_queue *txq); 3410 3411 static inline void netif_tx_schedule_all(struct net_device *dev) 3412 { 3413 unsigned int i; 3414 3415 for (i = 0; i < dev->num_tx_queues; i++) 3416 netif_schedule_queue(netdev_get_tx_queue(dev, i)); 3417 } 3418 3419 static __always_inline void netif_tx_start_queue(struct netdev_queue *dev_queue) 3420 { 3421 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state); 3422 } 3423 3424 /** 3425 * netif_start_queue - allow transmit 3426 * @dev: network device 3427 * 3428 * Allow upper layers to call the device hard_start_xmit routine. 3429 */ 3430 static inline void netif_start_queue(struct net_device *dev) 3431 { 3432 netif_tx_start_queue(netdev_get_tx_queue(dev, 0)); 3433 } 3434 3435 static inline void netif_tx_start_all_queues(struct net_device *dev) 3436 { 3437 unsigned int i; 3438 3439 for (i = 0; i < dev->num_tx_queues; i++) { 3440 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 3441 netif_tx_start_queue(txq); 3442 } 3443 } 3444 3445 void netif_tx_wake_queue(struct netdev_queue *dev_queue); 3446 3447 /** 3448 * netif_wake_queue - restart transmit 3449 * @dev: network device 3450 * 3451 * Allow upper layers to call the device hard_start_xmit routine. 3452 * Used for flow control when transmit resources are available. 3453 */ 3454 static inline void netif_wake_queue(struct net_device *dev) 3455 { 3456 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0)); 3457 } 3458 3459 static inline void netif_tx_wake_all_queues(struct net_device *dev) 3460 { 3461 unsigned int i; 3462 3463 for (i = 0; i < dev->num_tx_queues; i++) { 3464 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 3465 netif_tx_wake_queue(txq); 3466 } 3467 } 3468 3469 static __always_inline void netif_tx_stop_queue(struct netdev_queue *dev_queue) 3470 { 3471 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state); 3472 } 3473 3474 /** 3475 * netif_stop_queue - stop transmitted packets 3476 * @dev: network device 3477 * 3478 * Stop upper layers calling the device hard_start_xmit routine. 3479 * Used for flow control when transmit resources are unavailable. 3480 */ 3481 static inline void netif_stop_queue(struct net_device *dev) 3482 { 3483 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0)); 3484 } 3485 3486 void netif_tx_stop_all_queues(struct net_device *dev); 3487 3488 static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue) 3489 { 3490 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state); 3491 } 3492 3493 /** 3494 * netif_queue_stopped - test if transmit queue is flowblocked 3495 * @dev: network device 3496 * 3497 * Test if transmit queue on device is currently unable to send. 3498 */ 3499 static inline bool netif_queue_stopped(const struct net_device *dev) 3500 { 3501 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0)); 3502 } 3503 3504 static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue) 3505 { 3506 return dev_queue->state & QUEUE_STATE_ANY_XOFF; 3507 } 3508 3509 static inline bool 3510 netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue) 3511 { 3512 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN; 3513 } 3514 3515 static inline bool 3516 netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue) 3517 { 3518 return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN; 3519 } 3520 3521 /** 3522 * netdev_queue_set_dql_min_limit - set dql minimum limit 3523 * @dev_queue: pointer to transmit queue 3524 * @min_limit: dql minimum limit 3525 * 3526 * Forces xmit_more() to return true until the minimum threshold 3527 * defined by @min_limit is reached (or until the tx queue is 3528 * empty). Warning: to be use with care, misuse will impact the 3529 * latency. 3530 */ 3531 static inline void netdev_queue_set_dql_min_limit(struct netdev_queue *dev_queue, 3532 unsigned int min_limit) 3533 { 3534 #ifdef CONFIG_BQL 3535 dev_queue->dql.min_limit = min_limit; 3536 #endif 3537 } 3538 3539 /** 3540 * netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write 3541 * @dev_queue: pointer to transmit queue 3542 * 3543 * BQL enabled drivers might use this helper in their ndo_start_xmit(), 3544 * to give appropriate hint to the CPU. 3545 */ 3546 static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue) 3547 { 3548 #ifdef CONFIG_BQL 3549 prefetchw(&dev_queue->dql.num_queued); 3550 #endif 3551 } 3552 3553 /** 3554 * netdev_txq_bql_complete_prefetchw - prefetch bql data for write 3555 * @dev_queue: pointer to transmit queue 3556 * 3557 * BQL enabled drivers might use this helper in their TX completion path, 3558 * to give appropriate hint to the CPU. 3559 */ 3560 static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue) 3561 { 3562 #ifdef CONFIG_BQL 3563 prefetchw(&dev_queue->dql.limit); 3564 #endif 3565 } 3566 3567 static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue, 3568 unsigned int bytes) 3569 { 3570 #ifdef CONFIG_BQL 3571 dql_queued(&dev_queue->dql, bytes); 3572 3573 if (likely(dql_avail(&dev_queue->dql) >= 0)) 3574 return; 3575 3576 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state); 3577 3578 /* 3579 * The XOFF flag must be set before checking the dql_avail below, 3580 * because in netdev_tx_completed_queue we update the dql_completed 3581 * before checking the XOFF flag. 3582 */ 3583 smp_mb(); 3584 3585 /* check again in case another CPU has just made room avail */ 3586 if (unlikely(dql_avail(&dev_queue->dql) >= 0)) 3587 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state); 3588 #endif 3589 } 3590 3591 /* Variant of netdev_tx_sent_queue() for drivers that are aware 3592 * that they should not test BQL status themselves. 3593 * We do want to change __QUEUE_STATE_STACK_XOFF only for the last 3594 * skb of a batch. 3595 * Returns true if the doorbell must be used to kick the NIC. 3596 */ 3597 static inline bool __netdev_tx_sent_queue(struct netdev_queue *dev_queue, 3598 unsigned int bytes, 3599 bool xmit_more) 3600 { 3601 if (xmit_more) { 3602 #ifdef CONFIG_BQL 3603 dql_queued(&dev_queue->dql, bytes); 3604 #endif 3605 return netif_tx_queue_stopped(dev_queue); 3606 } 3607 netdev_tx_sent_queue(dev_queue, bytes); 3608 return true; 3609 } 3610 3611 /** 3612 * netdev_sent_queue - report the number of bytes queued to hardware 3613 * @dev: network device 3614 * @bytes: number of bytes queued to the hardware device queue 3615 * 3616 * Report the number of bytes queued for sending/completion to the network 3617 * device hardware queue. @bytes should be a good approximation and should 3618 * exactly match netdev_completed_queue() @bytes 3619 */ 3620 static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes) 3621 { 3622 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes); 3623 } 3624 3625 static inline bool __netdev_sent_queue(struct net_device *dev, 3626 unsigned int bytes, 3627 bool xmit_more) 3628 { 3629 return __netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes, 3630 xmit_more); 3631 } 3632 3633 static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue, 3634 unsigned int pkts, unsigned int bytes) 3635 { 3636 #ifdef CONFIG_BQL 3637 if (unlikely(!bytes)) 3638 return; 3639 3640 dql_completed(&dev_queue->dql, bytes); 3641 3642 /* 3643 * Without the memory barrier there is a small possiblity that 3644 * netdev_tx_sent_queue will miss the update and cause the queue to 3645 * be stopped forever 3646 */ 3647 smp_mb(); 3648 3649 if (unlikely(dql_avail(&dev_queue->dql) < 0)) 3650 return; 3651 3652 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state)) 3653 netif_schedule_queue(dev_queue); 3654 #endif 3655 } 3656 3657 /** 3658 * netdev_completed_queue - report bytes and packets completed by device 3659 * @dev: network device 3660 * @pkts: actual number of packets sent over the medium 3661 * @bytes: actual number of bytes sent over the medium 3662 * 3663 * Report the number of bytes and packets transmitted by the network device 3664 * hardware queue over the physical medium, @bytes must exactly match the 3665 * @bytes amount passed to netdev_sent_queue() 3666 */ 3667 static inline void netdev_completed_queue(struct net_device *dev, 3668 unsigned int pkts, unsigned int bytes) 3669 { 3670 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes); 3671 } 3672 3673 static inline void netdev_tx_reset_queue(struct netdev_queue *q) 3674 { 3675 #ifdef CONFIG_BQL 3676 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state); 3677 dql_reset(&q->dql); 3678 #endif 3679 } 3680 3681 /** 3682 * netdev_reset_queue - reset the packets and bytes count of a network device 3683 * @dev_queue: network device 3684 * 3685 * Reset the bytes and packet count of a network device and clear the 3686 * software flow control OFF bit for this network device 3687 */ 3688 static inline void netdev_reset_queue(struct net_device *dev_queue) 3689 { 3690 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0)); 3691 } 3692 3693 /** 3694 * netdev_cap_txqueue - check if selected tx queue exceeds device queues 3695 * @dev: network device 3696 * @queue_index: given tx queue index 3697 * 3698 * Returns 0 if given tx queue index >= number of device tx queues, 3699 * otherwise returns the originally passed tx queue index. 3700 */ 3701 static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index) 3702 { 3703 if (unlikely(queue_index >= dev->real_num_tx_queues)) { 3704 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n", 3705 dev->name, queue_index, 3706 dev->real_num_tx_queues); 3707 return 0; 3708 } 3709 3710 return queue_index; 3711 } 3712 3713 /** 3714 * netif_running - test if up 3715 * @dev: network device 3716 * 3717 * Test if the device has been brought up. 3718 */ 3719 static inline bool netif_running(const struct net_device *dev) 3720 { 3721 return test_bit(__LINK_STATE_START, &dev->state); 3722 } 3723 3724 /* 3725 * Routines to manage the subqueues on a device. We only need start, 3726 * stop, and a check if it's stopped. All other device management is 3727 * done at the overall netdevice level. 3728 * Also test the device if we're multiqueue. 3729 */ 3730 3731 /** 3732 * netif_start_subqueue - allow sending packets on subqueue 3733 * @dev: network device 3734 * @queue_index: sub queue index 3735 * 3736 * Start individual transmit queue of a device with multiple transmit queues. 3737 */ 3738 static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index) 3739 { 3740 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 3741 3742 netif_tx_start_queue(txq); 3743 } 3744 3745 /** 3746 * netif_stop_subqueue - stop sending packets on subqueue 3747 * @dev: network device 3748 * @queue_index: sub queue index 3749 * 3750 * Stop individual transmit queue of a device with multiple transmit queues. 3751 */ 3752 static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index) 3753 { 3754 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 3755 netif_tx_stop_queue(txq); 3756 } 3757 3758 /** 3759 * __netif_subqueue_stopped - test status of subqueue 3760 * @dev: network device 3761 * @queue_index: sub queue index 3762 * 3763 * Check individual transmit queue of a device with multiple transmit queues. 3764 */ 3765 static inline bool __netif_subqueue_stopped(const struct net_device *dev, 3766 u16 queue_index) 3767 { 3768 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 3769 3770 return netif_tx_queue_stopped(txq); 3771 } 3772 3773 /** 3774 * netif_subqueue_stopped - test status of subqueue 3775 * @dev: network device 3776 * @skb: sub queue buffer pointer 3777 * 3778 * Check individual transmit queue of a device with multiple transmit queues. 3779 */ 3780 static inline bool netif_subqueue_stopped(const struct net_device *dev, 3781 struct sk_buff *skb) 3782 { 3783 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb)); 3784 } 3785 3786 /** 3787 * netif_wake_subqueue - allow sending packets on subqueue 3788 * @dev: network device 3789 * @queue_index: sub queue index 3790 * 3791 * Resume individual transmit queue of a device with multiple transmit queues. 3792 */ 3793 static inline void netif_wake_subqueue(struct net_device *dev, u16 queue_index) 3794 { 3795 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); 3796 3797 netif_tx_wake_queue(txq); 3798 } 3799 3800 #ifdef CONFIG_XPS 3801 int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask, 3802 u16 index); 3803 int __netif_set_xps_queue(struct net_device *dev, const unsigned long *mask, 3804 u16 index, enum xps_map_type type); 3805 3806 /** 3807 * netif_attr_test_mask - Test a CPU or Rx queue set in a mask 3808 * @j: CPU/Rx queue index 3809 * @mask: bitmask of all cpus/rx queues 3810 * @nr_bits: number of bits in the bitmask 3811 * 3812 * Test if a CPU or Rx queue index is set in a mask of all CPU/Rx queues. 3813 */ 3814 static inline bool netif_attr_test_mask(unsigned long j, 3815 const unsigned long *mask, 3816 unsigned int nr_bits) 3817 { 3818 cpu_max_bits_warn(j, nr_bits); 3819 return test_bit(j, mask); 3820 } 3821 3822 /** 3823 * netif_attr_test_online - Test for online CPU/Rx queue 3824 * @j: CPU/Rx queue index 3825 * @online_mask: bitmask for CPUs/Rx queues that are online 3826 * @nr_bits: number of bits in the bitmask 3827 * 3828 * Returns true if a CPU/Rx queue is online. 3829 */ 3830 static inline bool netif_attr_test_online(unsigned long j, 3831 const unsigned long *online_mask, 3832 unsigned int nr_bits) 3833 { 3834 cpu_max_bits_warn(j, nr_bits); 3835 3836 if (online_mask) 3837 return test_bit(j, online_mask); 3838 3839 return (j < nr_bits); 3840 } 3841 3842 /** 3843 * netif_attrmask_next - get the next CPU/Rx queue in a cpu/Rx queues mask 3844 * @n: CPU/Rx queue index 3845 * @srcp: the cpumask/Rx queue mask pointer 3846 * @nr_bits: number of bits in the bitmask 3847 * 3848 * Returns >= nr_bits if no further CPUs/Rx queues set. 3849 */ 3850 static inline unsigned int netif_attrmask_next(int n, const unsigned long *srcp, 3851 unsigned int nr_bits) 3852 { 3853 /* -1 is a legal arg here. */ 3854 if (n != -1) 3855 cpu_max_bits_warn(n, nr_bits); 3856 3857 if (srcp) 3858 return find_next_bit(srcp, nr_bits, n + 1); 3859 3860 return n + 1; 3861 } 3862 3863 /** 3864 * netif_attrmask_next_and - get the next CPU/Rx queue in \*src1p & \*src2p 3865 * @n: CPU/Rx queue index 3866 * @src1p: the first CPUs/Rx queues mask pointer 3867 * @src2p: the second CPUs/Rx queues mask pointer 3868 * @nr_bits: number of bits in the bitmask 3869 * 3870 * Returns >= nr_bits if no further CPUs/Rx queues set in both. 3871 */ 3872 static inline int netif_attrmask_next_and(int n, const unsigned long *src1p, 3873 const unsigned long *src2p, 3874 unsigned int nr_bits) 3875 { 3876 /* -1 is a legal arg here. */ 3877 if (n != -1) 3878 cpu_max_bits_warn(n, nr_bits); 3879 3880 if (src1p && src2p) 3881 return find_next_and_bit(src1p, src2p, nr_bits, n + 1); 3882 else if (src1p) 3883 return find_next_bit(src1p, nr_bits, n + 1); 3884 else if (src2p) 3885 return find_next_bit(src2p, nr_bits, n + 1); 3886 3887 return n + 1; 3888 } 3889 #else 3890 static inline int netif_set_xps_queue(struct net_device *dev, 3891 const struct cpumask *mask, 3892 u16 index) 3893 { 3894 return 0; 3895 } 3896 3897 static inline int __netif_set_xps_queue(struct net_device *dev, 3898 const unsigned long *mask, 3899 u16 index, enum xps_map_type type) 3900 { 3901 return 0; 3902 } 3903 #endif 3904 3905 /** 3906 * netif_is_multiqueue - test if device has multiple transmit queues 3907 * @dev: network device 3908 * 3909 * Check if device has multiple transmit queues 3910 */ 3911 static inline bool netif_is_multiqueue(const struct net_device *dev) 3912 { 3913 return dev->num_tx_queues > 1; 3914 } 3915 3916 int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq); 3917 3918 #ifdef CONFIG_SYSFS 3919 int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq); 3920 #else 3921 static inline int netif_set_real_num_rx_queues(struct net_device *dev, 3922 unsigned int rxqs) 3923 { 3924 dev->real_num_rx_queues = rxqs; 3925 return 0; 3926 } 3927 #endif 3928 3929 static inline struct netdev_rx_queue * 3930 __netif_get_rx_queue(struct net_device *dev, unsigned int rxq) 3931 { 3932 return dev->_rx + rxq; 3933 } 3934 3935 #ifdef CONFIG_SYSFS 3936 static inline unsigned int get_netdev_rx_queue_index( 3937 struct netdev_rx_queue *queue) 3938 { 3939 struct net_device *dev = queue->dev; 3940 int index = queue - dev->_rx; 3941 3942 BUG_ON(index >= dev->num_rx_queues); 3943 return index; 3944 } 3945 #endif 3946 3947 #define DEFAULT_MAX_NUM_RSS_QUEUES (8) 3948 int netif_get_num_default_rss_queues(void); 3949 3950 enum skb_free_reason { 3951 SKB_REASON_CONSUMED, 3952 SKB_REASON_DROPPED, 3953 }; 3954 3955 void __dev_kfree_skb_irq(struct sk_buff *skb, enum skb_free_reason reason); 3956 void __dev_kfree_skb_any(struct sk_buff *skb, enum skb_free_reason reason); 3957 3958 /* 3959 * It is not allowed to call kfree_skb() or consume_skb() from hardware 3960 * interrupt context or with hardware interrupts being disabled. 3961 * (in_irq() || irqs_disabled()) 3962 * 3963 * We provide four helpers that can be used in following contexts : 3964 * 3965 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context, 3966 * replacing kfree_skb(skb) 3967 * 3968 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context. 3969 * Typically used in place of consume_skb(skb) in TX completion path 3970 * 3971 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context, 3972 * replacing kfree_skb(skb) 3973 * 3974 * dev_consume_skb_any(skb) when caller doesn't know its current irq context, 3975 * and consumed a packet. Used in place of consume_skb(skb) 3976 */ 3977 static inline void dev_kfree_skb_irq(struct sk_buff *skb) 3978 { 3979 __dev_kfree_skb_irq(skb, SKB_REASON_DROPPED); 3980 } 3981 3982 static inline void dev_consume_skb_irq(struct sk_buff *skb) 3983 { 3984 __dev_kfree_skb_irq(skb, SKB_REASON_CONSUMED); 3985 } 3986 3987 static inline void dev_kfree_skb_any(struct sk_buff *skb) 3988 { 3989 __dev_kfree_skb_any(skb, SKB_REASON_DROPPED); 3990 } 3991 3992 static inline void dev_consume_skb_any(struct sk_buff *skb) 3993 { 3994 __dev_kfree_skb_any(skb, SKB_REASON_CONSUMED); 3995 } 3996 3997 u32 bpf_prog_run_generic_xdp(struct sk_buff *skb, struct xdp_buff *xdp, 3998 struct bpf_prog *xdp_prog); 3999 void generic_xdp_tx(struct sk_buff *skb, struct bpf_prog *xdp_prog); 4000 int do_xdp_generic(struct bpf_prog *xdp_prog, struct sk_buff *skb); 4001 int netif_rx(struct sk_buff *skb); 4002 int netif_rx_ni(struct sk_buff *skb); 4003 int netif_rx_any_context(struct sk_buff *skb); 4004 int netif_receive_skb(struct sk_buff *skb); 4005 int netif_receive_skb_core(struct sk_buff *skb); 4006 void netif_receive_skb_list(struct list_head *head); 4007 gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb); 4008 void napi_gro_flush(struct napi_struct *napi, bool flush_old); 4009 struct sk_buff *napi_get_frags(struct napi_struct *napi); 4010 gro_result_t napi_gro_frags(struct napi_struct *napi); 4011 struct packet_offload *gro_find_receive_by_type(__be16 type); 4012 struct packet_offload *gro_find_complete_by_type(__be16 type); 4013 4014 static inline void napi_free_frags(struct napi_struct *napi) 4015 { 4016 kfree_skb(napi->skb); 4017 napi->skb = NULL; 4018 } 4019 4020 bool netdev_is_rx_handler_busy(struct net_device *dev); 4021 int netdev_rx_handler_register(struct net_device *dev, 4022 rx_handler_func_t *rx_handler, 4023 void *rx_handler_data); 4024 void netdev_rx_handler_unregister(struct net_device *dev); 4025 4026 bool dev_valid_name(const char *name); 4027 int get_user_ifreq(struct ifreq *ifr, void __user **ifrdata, void __user *arg); 4028 int put_user_ifreq(struct ifreq *ifr, void __user *arg); 4029 int dev_ioctl(struct net *net, unsigned int cmd, struct ifreq *ifr, 4030 void __user *data, bool *need_copyout); 4031 int dev_ifconf(struct net *net, struct ifconf __user *ifc); 4032 int dev_ethtool(struct net *net, struct ifreq *ifr, void __user *userdata); 4033 unsigned int dev_get_flags(const struct net_device *); 4034 int __dev_change_flags(struct net_device *dev, unsigned int flags, 4035 struct netlink_ext_ack *extack); 4036 int dev_change_flags(struct net_device *dev, unsigned int flags, 4037 struct netlink_ext_ack *extack); 4038 void __dev_notify_flags(struct net_device *, unsigned int old_flags, 4039 unsigned int gchanges); 4040 int dev_change_name(struct net_device *, const char *); 4041 int dev_set_alias(struct net_device *, const char *, size_t); 4042 int dev_get_alias(const struct net_device *, char *, size_t); 4043 int __dev_change_net_namespace(struct net_device *dev, struct net *net, 4044 const char *pat, int new_ifindex); 4045 static inline 4046 int dev_change_net_namespace(struct net_device *dev, struct net *net, 4047 const char *pat) 4048 { 4049 return __dev_change_net_namespace(dev, net, pat, 0); 4050 } 4051 int __dev_set_mtu(struct net_device *, int); 4052 int dev_validate_mtu(struct net_device *dev, int mtu, 4053 struct netlink_ext_ack *extack); 4054 int dev_set_mtu_ext(struct net_device *dev, int mtu, 4055 struct netlink_ext_ack *extack); 4056 int dev_set_mtu(struct net_device *, int); 4057 int dev_change_tx_queue_len(struct net_device *, unsigned long); 4058 void dev_set_group(struct net_device *, int); 4059 int dev_pre_changeaddr_notify(struct net_device *dev, const char *addr, 4060 struct netlink_ext_ack *extack); 4061 int dev_set_mac_address(struct net_device *dev, struct sockaddr *sa, 4062 struct netlink_ext_ack *extack); 4063 int dev_set_mac_address_user(struct net_device *dev, struct sockaddr *sa, 4064 struct netlink_ext_ack *extack); 4065 int dev_get_mac_address(struct sockaddr *sa, struct net *net, char *dev_name); 4066 int dev_change_carrier(struct net_device *, bool new_carrier); 4067 int dev_get_phys_port_id(struct net_device *dev, 4068 struct netdev_phys_item_id *ppid); 4069 int dev_get_phys_port_name(struct net_device *dev, 4070 char *name, size_t len); 4071 int dev_get_port_parent_id(struct net_device *dev, 4072 struct netdev_phys_item_id *ppid, bool recurse); 4073 bool netdev_port_same_parent_id(struct net_device *a, struct net_device *b); 4074 int dev_change_proto_down(struct net_device *dev, bool proto_down); 4075 int dev_change_proto_down_generic(struct net_device *dev, bool proto_down); 4076 void dev_change_proto_down_reason(struct net_device *dev, unsigned long mask, 4077 u32 value); 4078 struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev, bool *again); 4079 struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev, 4080 struct netdev_queue *txq, int *ret); 4081 4082 typedef int (*bpf_op_t)(struct net_device *dev, struct netdev_bpf *bpf); 4083 int dev_change_xdp_fd(struct net_device *dev, struct netlink_ext_ack *extack, 4084 int fd, int expected_fd, u32 flags); 4085 int bpf_xdp_link_attach(const union bpf_attr *attr, struct bpf_prog *prog); 4086 u32 dev_xdp_prog_id(struct net_device *dev, enum bpf_xdp_mode mode); 4087 4088 int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb); 4089 int dev_forward_skb(struct net_device *dev, struct sk_buff *skb); 4090 int dev_forward_skb_nomtu(struct net_device *dev, struct sk_buff *skb); 4091 bool is_skb_forwardable(const struct net_device *dev, 4092 const struct sk_buff *skb); 4093 4094 static __always_inline bool __is_skb_forwardable(const struct net_device *dev, 4095 const struct sk_buff *skb, 4096 const bool check_mtu) 4097 { 4098 const u32 vlan_hdr_len = 4; /* VLAN_HLEN */ 4099 unsigned int len; 4100 4101 if (!(dev->flags & IFF_UP)) 4102 return false; 4103 4104 if (!check_mtu) 4105 return true; 4106 4107 len = dev->mtu + dev->hard_header_len + vlan_hdr_len; 4108 if (skb->len <= len) 4109 return true; 4110 4111 /* if TSO is enabled, we don't care about the length as the packet 4112 * could be forwarded without being segmented before 4113 */ 4114 if (skb_is_gso(skb)) 4115 return true; 4116 4117 return false; 4118 } 4119 4120 static __always_inline int ____dev_forward_skb(struct net_device *dev, 4121 struct sk_buff *skb, 4122 const bool check_mtu) 4123 { 4124 if (skb_orphan_frags(skb, GFP_ATOMIC) || 4125 unlikely(!__is_skb_forwardable(dev, skb, check_mtu))) { 4126 atomic_long_inc(&dev->rx_dropped); 4127 kfree_skb(skb); 4128 return NET_RX_DROP; 4129 } 4130 4131 skb_scrub_packet(skb, !net_eq(dev_net(dev), dev_net(skb->dev))); 4132 skb->priority = 0; 4133 return 0; 4134 } 4135 4136 bool dev_nit_active(struct net_device *dev); 4137 void dev_queue_xmit_nit(struct sk_buff *skb, struct net_device *dev); 4138 4139 extern int netdev_budget; 4140 extern unsigned int netdev_budget_usecs; 4141 4142 /* Called by rtnetlink.c:rtnl_unlock() */ 4143 void netdev_run_todo(void); 4144 4145 /** 4146 * dev_put - release reference to device 4147 * @dev: network device 4148 * 4149 * Release reference to device to allow it to be freed. 4150 */ 4151 static inline void dev_put(struct net_device *dev) 4152 { 4153 #ifdef CONFIG_PCPU_DEV_REFCNT 4154 this_cpu_dec(*dev->pcpu_refcnt); 4155 #else 4156 refcount_dec(&dev->dev_refcnt); 4157 #endif 4158 } 4159 4160 /** 4161 * dev_hold - get reference to device 4162 * @dev: network device 4163 * 4164 * Hold reference to device to keep it from being freed. 4165 */ 4166 static inline void dev_hold(struct net_device *dev) 4167 { 4168 #ifdef CONFIG_PCPU_DEV_REFCNT 4169 this_cpu_inc(*dev->pcpu_refcnt); 4170 #else 4171 refcount_inc(&dev->dev_refcnt); 4172 #endif 4173 } 4174 4175 /* Carrier loss detection, dial on demand. The functions netif_carrier_on 4176 * and _off may be called from IRQ context, but it is caller 4177 * who is responsible for serialization of these calls. 4178 * 4179 * The name carrier is inappropriate, these functions should really be 4180 * called netif_lowerlayer_*() because they represent the state of any 4181 * kind of lower layer not just hardware media. 4182 */ 4183 4184 void linkwatch_init_dev(struct net_device *dev); 4185 void linkwatch_fire_event(struct net_device *dev); 4186 void linkwatch_forget_dev(struct net_device *dev); 4187 4188 /** 4189 * netif_carrier_ok - test if carrier present 4190 * @dev: network device 4191 * 4192 * Check if carrier is present on device 4193 */ 4194 static inline bool netif_carrier_ok(const struct net_device *dev) 4195 { 4196 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state); 4197 } 4198 4199 unsigned long dev_trans_start(struct net_device *dev); 4200 4201 void __netdev_watchdog_up(struct net_device *dev); 4202 4203 void netif_carrier_on(struct net_device *dev); 4204 void netif_carrier_off(struct net_device *dev); 4205 void netif_carrier_event(struct net_device *dev); 4206 4207 /** 4208 * netif_dormant_on - mark device as dormant. 4209 * @dev: network device 4210 * 4211 * Mark device as dormant (as per RFC2863). 4212 * 4213 * The dormant state indicates that the relevant interface is not 4214 * actually in a condition to pass packets (i.e., it is not 'up') but is 4215 * in a "pending" state, waiting for some external event. For "on- 4216 * demand" interfaces, this new state identifies the situation where the 4217 * interface is waiting for events to place it in the up state. 4218 */ 4219 static inline void netif_dormant_on(struct net_device *dev) 4220 { 4221 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state)) 4222 linkwatch_fire_event(dev); 4223 } 4224 4225 /** 4226 * netif_dormant_off - set device as not dormant. 4227 * @dev: network device 4228 * 4229 * Device is not in dormant state. 4230 */ 4231 static inline void netif_dormant_off(struct net_device *dev) 4232 { 4233 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state)) 4234 linkwatch_fire_event(dev); 4235 } 4236 4237 /** 4238 * netif_dormant - test if device is dormant 4239 * @dev: network device 4240 * 4241 * Check if device is dormant. 4242 */ 4243 static inline bool netif_dormant(const struct net_device *dev) 4244 { 4245 return test_bit(__LINK_STATE_DORMANT, &dev->state); 4246 } 4247 4248 4249 /** 4250 * netif_testing_on - mark device as under test. 4251 * @dev: network device 4252 * 4253 * Mark device as under test (as per RFC2863). 4254 * 4255 * The testing state indicates that some test(s) must be performed on 4256 * the interface. After completion, of the test, the interface state 4257 * will change to up, dormant, or down, as appropriate. 4258 */ 4259 static inline void netif_testing_on(struct net_device *dev) 4260 { 4261 if (!test_and_set_bit(__LINK_STATE_TESTING, &dev->state)) 4262 linkwatch_fire_event(dev); 4263 } 4264 4265 /** 4266 * netif_testing_off - set device as not under test. 4267 * @dev: network device 4268 * 4269 * Device is not in testing state. 4270 */ 4271 static inline void netif_testing_off(struct net_device *dev) 4272 { 4273 if (test_and_clear_bit(__LINK_STATE_TESTING, &dev->state)) 4274 linkwatch_fire_event(dev); 4275 } 4276 4277 /** 4278 * netif_testing - test if device is under test 4279 * @dev: network device 4280 * 4281 * Check if device is under test 4282 */ 4283 static inline bool netif_testing(const struct net_device *dev) 4284 { 4285 return test_bit(__LINK_STATE_TESTING, &dev->state); 4286 } 4287 4288 4289 /** 4290 * netif_oper_up - test if device is operational 4291 * @dev: network device 4292 * 4293 * Check if carrier is operational 4294 */ 4295 static inline bool netif_oper_up(const struct net_device *dev) 4296 { 4297 return (dev->operstate == IF_OPER_UP || 4298 dev->operstate == IF_OPER_UNKNOWN /* backward compat */); 4299 } 4300 4301 /** 4302 * netif_device_present - is device available or removed 4303 * @dev: network device 4304 * 4305 * Check if device has not been removed from system. 4306 */ 4307 static inline bool netif_device_present(const struct net_device *dev) 4308 { 4309 return test_bit(__LINK_STATE_PRESENT, &dev->state); 4310 } 4311 4312 void netif_device_detach(struct net_device *dev); 4313 4314 void netif_device_attach(struct net_device *dev); 4315 4316 /* 4317 * Network interface message level settings 4318 */ 4319 4320 enum { 4321 NETIF_MSG_DRV_BIT, 4322 NETIF_MSG_PROBE_BIT, 4323 NETIF_MSG_LINK_BIT, 4324 NETIF_MSG_TIMER_BIT, 4325 NETIF_MSG_IFDOWN_BIT, 4326 NETIF_MSG_IFUP_BIT, 4327 NETIF_MSG_RX_ERR_BIT, 4328 NETIF_MSG_TX_ERR_BIT, 4329 NETIF_MSG_TX_QUEUED_BIT, 4330 NETIF_MSG_INTR_BIT, 4331 NETIF_MSG_TX_DONE_BIT, 4332 NETIF_MSG_RX_STATUS_BIT, 4333 NETIF_MSG_PKTDATA_BIT, 4334 NETIF_MSG_HW_BIT, 4335 NETIF_MSG_WOL_BIT, 4336 4337 /* When you add a new bit above, update netif_msg_class_names array 4338 * in net/ethtool/common.c 4339 */ 4340 NETIF_MSG_CLASS_COUNT, 4341 }; 4342 /* Both ethtool_ops interface and internal driver implementation use u32 */ 4343 static_assert(NETIF_MSG_CLASS_COUNT <= 32); 4344 4345 #define __NETIF_MSG_BIT(bit) ((u32)1 << (bit)) 4346 #define __NETIF_MSG(name) __NETIF_MSG_BIT(NETIF_MSG_ ## name ## _BIT) 4347 4348 #define NETIF_MSG_DRV __NETIF_MSG(DRV) 4349 #define NETIF_MSG_PROBE __NETIF_MSG(PROBE) 4350 #define NETIF_MSG_LINK __NETIF_MSG(LINK) 4351 #define NETIF_MSG_TIMER __NETIF_MSG(TIMER) 4352 #define NETIF_MSG_IFDOWN __NETIF_MSG(IFDOWN) 4353 #define NETIF_MSG_IFUP __NETIF_MSG(IFUP) 4354 #define NETIF_MSG_RX_ERR __NETIF_MSG(RX_ERR) 4355 #define NETIF_MSG_TX_ERR __NETIF_MSG(TX_ERR) 4356 #define NETIF_MSG_TX_QUEUED __NETIF_MSG(TX_QUEUED) 4357 #define NETIF_MSG_INTR __NETIF_MSG(INTR) 4358 #define NETIF_MSG_TX_DONE __NETIF_MSG(TX_DONE) 4359 #define NETIF_MSG_RX_STATUS __NETIF_MSG(RX_STATUS) 4360 #define NETIF_MSG_PKTDATA __NETIF_MSG(PKTDATA) 4361 #define NETIF_MSG_HW __NETIF_MSG(HW) 4362 #define NETIF_MSG_WOL __NETIF_MSG(WOL) 4363 4364 #define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV) 4365 #define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE) 4366 #define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK) 4367 #define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER) 4368 #define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN) 4369 #define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP) 4370 #define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR) 4371 #define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR) 4372 #define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED) 4373 #define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR) 4374 #define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE) 4375 #define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS) 4376 #define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA) 4377 #define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW) 4378 #define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL) 4379 4380 static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits) 4381 { 4382 /* use default */ 4383 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8)) 4384 return default_msg_enable_bits; 4385 if (debug_value == 0) /* no output */ 4386 return 0; 4387 /* set low N bits */ 4388 return (1U << debug_value) - 1; 4389 } 4390 4391 static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu) 4392 { 4393 spin_lock(&txq->_xmit_lock); 4394 txq->xmit_lock_owner = cpu; 4395 } 4396 4397 static inline bool __netif_tx_acquire(struct netdev_queue *txq) 4398 { 4399 __acquire(&txq->_xmit_lock); 4400 return true; 4401 } 4402 4403 static inline void __netif_tx_release(struct netdev_queue *txq) 4404 { 4405 __release(&txq->_xmit_lock); 4406 } 4407 4408 static inline void __netif_tx_lock_bh(struct netdev_queue *txq) 4409 { 4410 spin_lock_bh(&txq->_xmit_lock); 4411 txq->xmit_lock_owner = smp_processor_id(); 4412 } 4413 4414 static inline bool __netif_tx_trylock(struct netdev_queue *txq) 4415 { 4416 bool ok = spin_trylock(&txq->_xmit_lock); 4417 if (likely(ok)) 4418 txq->xmit_lock_owner = smp_processor_id(); 4419 return ok; 4420 } 4421 4422 static inline void __netif_tx_unlock(struct netdev_queue *txq) 4423 { 4424 txq->xmit_lock_owner = -1; 4425 spin_unlock(&txq->_xmit_lock); 4426 } 4427 4428 static inline void __netif_tx_unlock_bh(struct netdev_queue *txq) 4429 { 4430 txq->xmit_lock_owner = -1; 4431 spin_unlock_bh(&txq->_xmit_lock); 4432 } 4433 4434 static inline void txq_trans_update(struct netdev_queue *txq) 4435 { 4436 if (txq->xmit_lock_owner != -1) 4437 txq->trans_start = jiffies; 4438 } 4439 4440 /* legacy drivers only, netdev_start_xmit() sets txq->trans_start */ 4441 static inline void netif_trans_update(struct net_device *dev) 4442 { 4443 struct netdev_queue *txq = netdev_get_tx_queue(dev, 0); 4444 4445 if (txq->trans_start != jiffies) 4446 txq->trans_start = jiffies; 4447 } 4448 4449 /** 4450 * netif_tx_lock - grab network device transmit lock 4451 * @dev: network device 4452 * 4453 * Get network device transmit lock 4454 */ 4455 static inline void netif_tx_lock(struct net_device *dev) 4456 { 4457 unsigned int i; 4458 int cpu; 4459 4460 spin_lock(&dev->tx_global_lock); 4461 cpu = smp_processor_id(); 4462 for (i = 0; i < dev->num_tx_queues; i++) { 4463 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 4464 4465 /* We are the only thread of execution doing a 4466 * freeze, but we have to grab the _xmit_lock in 4467 * order to synchronize with threads which are in 4468 * the ->hard_start_xmit() handler and already 4469 * checked the frozen bit. 4470 */ 4471 __netif_tx_lock(txq, cpu); 4472 set_bit(__QUEUE_STATE_FROZEN, &txq->state); 4473 __netif_tx_unlock(txq); 4474 } 4475 } 4476 4477 static inline void netif_tx_lock_bh(struct net_device *dev) 4478 { 4479 local_bh_disable(); 4480 netif_tx_lock(dev); 4481 } 4482 4483 static inline void netif_tx_unlock(struct net_device *dev) 4484 { 4485 unsigned int i; 4486 4487 for (i = 0; i < dev->num_tx_queues; i++) { 4488 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 4489 4490 /* No need to grab the _xmit_lock here. If the 4491 * queue is not stopped for another reason, we 4492 * force a schedule. 4493 */ 4494 clear_bit(__QUEUE_STATE_FROZEN, &txq->state); 4495 netif_schedule_queue(txq); 4496 } 4497 spin_unlock(&dev->tx_global_lock); 4498 } 4499 4500 static inline void netif_tx_unlock_bh(struct net_device *dev) 4501 { 4502 netif_tx_unlock(dev); 4503 local_bh_enable(); 4504 } 4505 4506 #define HARD_TX_LOCK(dev, txq, cpu) { \ 4507 if ((dev->features & NETIF_F_LLTX) == 0) { \ 4508 __netif_tx_lock(txq, cpu); \ 4509 } else { \ 4510 __netif_tx_acquire(txq); \ 4511 } \ 4512 } 4513 4514 #define HARD_TX_TRYLOCK(dev, txq) \ 4515 (((dev->features & NETIF_F_LLTX) == 0) ? \ 4516 __netif_tx_trylock(txq) : \ 4517 __netif_tx_acquire(txq)) 4518 4519 #define HARD_TX_UNLOCK(dev, txq) { \ 4520 if ((dev->features & NETIF_F_LLTX) == 0) { \ 4521 __netif_tx_unlock(txq); \ 4522 } else { \ 4523 __netif_tx_release(txq); \ 4524 } \ 4525 } 4526 4527 static inline void netif_tx_disable(struct net_device *dev) 4528 { 4529 unsigned int i; 4530 int cpu; 4531 4532 local_bh_disable(); 4533 cpu = smp_processor_id(); 4534 spin_lock(&dev->tx_global_lock); 4535 for (i = 0; i < dev->num_tx_queues; i++) { 4536 struct netdev_queue *txq = netdev_get_tx_queue(dev, i); 4537 4538 __netif_tx_lock(txq, cpu); 4539 netif_tx_stop_queue(txq); 4540 __netif_tx_unlock(txq); 4541 } 4542 spin_unlock(&dev->tx_global_lock); 4543 local_bh_enable(); 4544 } 4545 4546 static inline void netif_addr_lock(struct net_device *dev) 4547 { 4548 unsigned char nest_level = 0; 4549 4550 #ifdef CONFIG_LOCKDEP 4551 nest_level = dev->nested_level; 4552 #endif 4553 spin_lock_nested(&dev->addr_list_lock, nest_level); 4554 } 4555 4556 static inline void netif_addr_lock_bh(struct net_device *dev) 4557 { 4558 unsigned char nest_level = 0; 4559 4560 #ifdef CONFIG_LOCKDEP 4561 nest_level = dev->nested_level; 4562 #endif 4563 local_bh_disable(); 4564 spin_lock_nested(&dev->addr_list_lock, nest_level); 4565 } 4566 4567 static inline void netif_addr_unlock(struct net_device *dev) 4568 { 4569 spin_unlock(&dev->addr_list_lock); 4570 } 4571 4572 static inline void netif_addr_unlock_bh(struct net_device *dev) 4573 { 4574 spin_unlock_bh(&dev->addr_list_lock); 4575 } 4576 4577 /* 4578 * dev_addrs walker. Should be used only for read access. Call with 4579 * rcu_read_lock held. 4580 */ 4581 #define for_each_dev_addr(dev, ha) \ 4582 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list) 4583 4584 /* These functions live elsewhere (drivers/net/net_init.c, but related) */ 4585 4586 void ether_setup(struct net_device *dev); 4587 4588 /* Support for loadable net-drivers */ 4589 struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name, 4590 unsigned char name_assign_type, 4591 void (*setup)(struct net_device *), 4592 unsigned int txqs, unsigned int rxqs); 4593 #define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \ 4594 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1) 4595 4596 #define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \ 4597 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \ 4598 count) 4599 4600 int register_netdev(struct net_device *dev); 4601 void unregister_netdev(struct net_device *dev); 4602 4603 int devm_register_netdev(struct device *dev, struct net_device *ndev); 4604 4605 /* General hardware address lists handling functions */ 4606 int __hw_addr_sync(struct netdev_hw_addr_list *to_list, 4607 struct netdev_hw_addr_list *from_list, int addr_len); 4608 void __hw_addr_unsync(struct netdev_hw_addr_list *to_list, 4609 struct netdev_hw_addr_list *from_list, int addr_len); 4610 int __hw_addr_sync_dev(struct netdev_hw_addr_list *list, 4611 struct net_device *dev, 4612 int (*sync)(struct net_device *, const unsigned char *), 4613 int (*unsync)(struct net_device *, 4614 const unsigned char *)); 4615 int __hw_addr_ref_sync_dev(struct netdev_hw_addr_list *list, 4616 struct net_device *dev, 4617 int (*sync)(struct net_device *, 4618 const unsigned char *, int), 4619 int (*unsync)(struct net_device *, 4620 const unsigned char *, int)); 4621 void __hw_addr_ref_unsync_dev(struct netdev_hw_addr_list *list, 4622 struct net_device *dev, 4623 int (*unsync)(struct net_device *, 4624 const unsigned char *, int)); 4625 void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list, 4626 struct net_device *dev, 4627 int (*unsync)(struct net_device *, 4628 const unsigned char *)); 4629 void __hw_addr_init(struct netdev_hw_addr_list *list); 4630 4631 /* Functions used for device addresses handling */ 4632 int dev_addr_add(struct net_device *dev, const unsigned char *addr, 4633 unsigned char addr_type); 4634 int dev_addr_del(struct net_device *dev, const unsigned char *addr, 4635 unsigned char addr_type); 4636 void dev_addr_flush(struct net_device *dev); 4637 int dev_addr_init(struct net_device *dev); 4638 4639 /* Functions used for unicast addresses handling */ 4640 int dev_uc_add(struct net_device *dev, const unsigned char *addr); 4641 int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr); 4642 int dev_uc_del(struct net_device *dev, const unsigned char *addr); 4643 int dev_uc_sync(struct net_device *to, struct net_device *from); 4644 int dev_uc_sync_multiple(struct net_device *to, struct net_device *from); 4645 void dev_uc_unsync(struct net_device *to, struct net_device *from); 4646 void dev_uc_flush(struct net_device *dev); 4647 void dev_uc_init(struct net_device *dev); 4648 4649 /** 4650 * __dev_uc_sync - Synchonize device's unicast list 4651 * @dev: device to sync 4652 * @sync: function to call if address should be added 4653 * @unsync: function to call if address should be removed 4654 * 4655 * Add newly added addresses to the interface, and release 4656 * addresses that have been deleted. 4657 */ 4658 static inline int __dev_uc_sync(struct net_device *dev, 4659 int (*sync)(struct net_device *, 4660 const unsigned char *), 4661 int (*unsync)(struct net_device *, 4662 const unsigned char *)) 4663 { 4664 return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync); 4665 } 4666 4667 /** 4668 * __dev_uc_unsync - Remove synchronized addresses from device 4669 * @dev: device to sync 4670 * @unsync: function to call if address should be removed 4671 * 4672 * Remove all addresses that were added to the device by dev_uc_sync(). 4673 */ 4674 static inline void __dev_uc_unsync(struct net_device *dev, 4675 int (*unsync)(struct net_device *, 4676 const unsigned char *)) 4677 { 4678 __hw_addr_unsync_dev(&dev->uc, dev, unsync); 4679 } 4680 4681 /* Functions used for multicast addresses handling */ 4682 int dev_mc_add(struct net_device *dev, const unsigned char *addr); 4683 int dev_mc_add_global(struct net_device *dev, const unsigned char *addr); 4684 int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr); 4685 int dev_mc_del(struct net_device *dev, const unsigned char *addr); 4686 int dev_mc_del_global(struct net_device *dev, const unsigned char *addr); 4687 int dev_mc_sync(struct net_device *to, struct net_device *from); 4688 int dev_mc_sync_multiple(struct net_device *to, struct net_device *from); 4689 void dev_mc_unsync(struct net_device *to, struct net_device *from); 4690 void dev_mc_flush(struct net_device *dev); 4691 void dev_mc_init(struct net_device *dev); 4692 4693 /** 4694 * __dev_mc_sync - Synchonize device's multicast list 4695 * @dev: device to sync 4696 * @sync: function to call if address should be added 4697 * @unsync: function to call if address should be removed 4698 * 4699 * Add newly added addresses to the interface, and release 4700 * addresses that have been deleted. 4701 */ 4702 static inline int __dev_mc_sync(struct net_device *dev, 4703 int (*sync)(struct net_device *, 4704 const unsigned char *), 4705 int (*unsync)(struct net_device *, 4706 const unsigned char *)) 4707 { 4708 return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync); 4709 } 4710 4711 /** 4712 * __dev_mc_unsync - Remove synchronized addresses from device 4713 * @dev: device to sync 4714 * @unsync: function to call if address should be removed 4715 * 4716 * Remove all addresses that were added to the device by dev_mc_sync(). 4717 */ 4718 static inline void __dev_mc_unsync(struct net_device *dev, 4719 int (*unsync)(struct net_device *, 4720 const unsigned char *)) 4721 { 4722 __hw_addr_unsync_dev(&dev->mc, dev, unsync); 4723 } 4724 4725 /* Functions used for secondary unicast and multicast support */ 4726 void dev_set_rx_mode(struct net_device *dev); 4727 void __dev_set_rx_mode(struct net_device *dev); 4728 int dev_set_promiscuity(struct net_device *dev, int inc); 4729 int dev_set_allmulti(struct net_device *dev, int inc); 4730 void netdev_state_change(struct net_device *dev); 4731 void __netdev_notify_peers(struct net_device *dev); 4732 void netdev_notify_peers(struct net_device *dev); 4733 void netdev_features_change(struct net_device *dev); 4734 /* Load a device via the kmod */ 4735 void dev_load(struct net *net, const char *name); 4736 struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev, 4737 struct rtnl_link_stats64 *storage); 4738 void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64, 4739 const struct net_device_stats *netdev_stats); 4740 void dev_fetch_sw_netstats(struct rtnl_link_stats64 *s, 4741 const struct pcpu_sw_netstats __percpu *netstats); 4742 void dev_get_tstats64(struct net_device *dev, struct rtnl_link_stats64 *s); 4743 4744 extern int netdev_max_backlog; 4745 extern int netdev_tstamp_prequeue; 4746 extern int netdev_unregister_timeout_secs; 4747 extern int weight_p; 4748 extern int dev_weight_rx_bias; 4749 extern int dev_weight_tx_bias; 4750 extern int dev_rx_weight; 4751 extern int dev_tx_weight; 4752 extern int gro_normal_batch; 4753 4754 enum { 4755 NESTED_SYNC_IMM_BIT, 4756 NESTED_SYNC_TODO_BIT, 4757 }; 4758 4759 #define __NESTED_SYNC_BIT(bit) ((u32)1 << (bit)) 4760 #define __NESTED_SYNC(name) __NESTED_SYNC_BIT(NESTED_SYNC_ ## name ## _BIT) 4761 4762 #define NESTED_SYNC_IMM __NESTED_SYNC(IMM) 4763 #define NESTED_SYNC_TODO __NESTED_SYNC(TODO) 4764 4765 struct netdev_nested_priv { 4766 unsigned char flags; 4767 void *data; 4768 }; 4769 4770 bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev); 4771 struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev, 4772 struct list_head **iter); 4773 struct net_device *netdev_all_upper_get_next_dev_rcu(struct net_device *dev, 4774 struct list_head **iter); 4775 4776 #ifdef CONFIG_LOCKDEP 4777 static LIST_HEAD(net_unlink_list); 4778 4779 static inline void net_unlink_todo(struct net_device *dev) 4780 { 4781 if (list_empty(&dev->unlink_list)) 4782 list_add_tail(&dev->unlink_list, &net_unlink_list); 4783 } 4784 #endif 4785 4786 /* iterate through upper list, must be called under RCU read lock */ 4787 #define netdev_for_each_upper_dev_rcu(dev, updev, iter) \ 4788 for (iter = &(dev)->adj_list.upper, \ 4789 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \ 4790 updev; \ 4791 updev = netdev_upper_get_next_dev_rcu(dev, &(iter))) 4792 4793 int netdev_walk_all_upper_dev_rcu(struct net_device *dev, 4794 int (*fn)(struct net_device *upper_dev, 4795 struct netdev_nested_priv *priv), 4796 struct netdev_nested_priv *priv); 4797 4798 bool netdev_has_upper_dev_all_rcu(struct net_device *dev, 4799 struct net_device *upper_dev); 4800 4801 bool netdev_has_any_upper_dev(struct net_device *dev); 4802 4803 void *netdev_lower_get_next_private(struct net_device *dev, 4804 struct list_head **iter); 4805 void *netdev_lower_get_next_private_rcu(struct net_device *dev, 4806 struct list_head **iter); 4807 4808 #define netdev_for_each_lower_private(dev, priv, iter) \ 4809 for (iter = (dev)->adj_list.lower.next, \ 4810 priv = netdev_lower_get_next_private(dev, &(iter)); \ 4811 priv; \ 4812 priv = netdev_lower_get_next_private(dev, &(iter))) 4813 4814 #define netdev_for_each_lower_private_rcu(dev, priv, iter) \ 4815 for (iter = &(dev)->adj_list.lower, \ 4816 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \ 4817 priv; \ 4818 priv = netdev_lower_get_next_private_rcu(dev, &(iter))) 4819 4820 void *netdev_lower_get_next(struct net_device *dev, 4821 struct list_head **iter); 4822 4823 #define netdev_for_each_lower_dev(dev, ldev, iter) \ 4824 for (iter = (dev)->adj_list.lower.next, \ 4825 ldev = netdev_lower_get_next(dev, &(iter)); \ 4826 ldev; \ 4827 ldev = netdev_lower_get_next(dev, &(iter))) 4828 4829 struct net_device *netdev_next_lower_dev_rcu(struct net_device *dev, 4830 struct list_head **iter); 4831 int netdev_walk_all_lower_dev(struct net_device *dev, 4832 int (*fn)(struct net_device *lower_dev, 4833 struct netdev_nested_priv *priv), 4834 struct netdev_nested_priv *priv); 4835 int netdev_walk_all_lower_dev_rcu(struct net_device *dev, 4836 int (*fn)(struct net_device *lower_dev, 4837 struct netdev_nested_priv *priv), 4838 struct netdev_nested_priv *priv); 4839 4840 void *netdev_adjacent_get_private(struct list_head *adj_list); 4841 void *netdev_lower_get_first_private_rcu(struct net_device *dev); 4842 struct net_device *netdev_master_upper_dev_get(struct net_device *dev); 4843 struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev); 4844 int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev, 4845 struct netlink_ext_ack *extack); 4846 int netdev_master_upper_dev_link(struct net_device *dev, 4847 struct net_device *upper_dev, 4848 void *upper_priv, void *upper_info, 4849 struct netlink_ext_ack *extack); 4850 void netdev_upper_dev_unlink(struct net_device *dev, 4851 struct net_device *upper_dev); 4852 int netdev_adjacent_change_prepare(struct net_device *old_dev, 4853 struct net_device *new_dev, 4854 struct net_device *dev, 4855 struct netlink_ext_ack *extack); 4856 void netdev_adjacent_change_commit(struct net_device *old_dev, 4857 struct net_device *new_dev, 4858 struct net_device *dev); 4859 void netdev_adjacent_change_abort(struct net_device *old_dev, 4860 struct net_device *new_dev, 4861 struct net_device *dev); 4862 void netdev_adjacent_rename_links(struct net_device *dev, char *oldname); 4863 void *netdev_lower_dev_get_private(struct net_device *dev, 4864 struct net_device *lower_dev); 4865 void netdev_lower_state_changed(struct net_device *lower_dev, 4866 void *lower_state_info); 4867 4868 /* RSS keys are 40 or 52 bytes long */ 4869 #define NETDEV_RSS_KEY_LEN 52 4870 extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN] __read_mostly; 4871 void netdev_rss_key_fill(void *buffer, size_t len); 4872 4873 int skb_checksum_help(struct sk_buff *skb); 4874 int skb_crc32c_csum_help(struct sk_buff *skb); 4875 int skb_csum_hwoffload_help(struct sk_buff *skb, 4876 const netdev_features_t features); 4877 4878 struct sk_buff *__skb_gso_segment(struct sk_buff *skb, 4879 netdev_features_t features, bool tx_path); 4880 struct sk_buff *skb_mac_gso_segment(struct sk_buff *skb, 4881 netdev_features_t features); 4882 4883 struct netdev_bonding_info { 4884 ifslave slave; 4885 ifbond master; 4886 }; 4887 4888 struct netdev_notifier_bonding_info { 4889 struct netdev_notifier_info info; /* must be first */ 4890 struct netdev_bonding_info bonding_info; 4891 }; 4892 4893 void netdev_bonding_info_change(struct net_device *dev, 4894 struct netdev_bonding_info *bonding_info); 4895 4896 #if IS_ENABLED(CONFIG_ETHTOOL_NETLINK) 4897 void ethtool_notify(struct net_device *dev, unsigned int cmd, const void *data); 4898 #else 4899 static inline void ethtool_notify(struct net_device *dev, unsigned int cmd, 4900 const void *data) 4901 { 4902 } 4903 #endif 4904 4905 static inline 4906 struct sk_buff *skb_gso_segment(struct sk_buff *skb, netdev_features_t features) 4907 { 4908 return __skb_gso_segment(skb, features, true); 4909 } 4910 __be16 skb_network_protocol(struct sk_buff *skb, int *depth); 4911 4912 static inline bool can_checksum_protocol(netdev_features_t features, 4913 __be16 protocol) 4914 { 4915 if (protocol == htons(ETH_P_FCOE)) 4916 return !!(features & NETIF_F_FCOE_CRC); 4917 4918 /* Assume this is an IP checksum (not SCTP CRC) */ 4919 4920 if (features & NETIF_F_HW_CSUM) { 4921 /* Can checksum everything */ 4922 return true; 4923 } 4924 4925 switch (protocol) { 4926 case htons(ETH_P_IP): 4927 return !!(features & NETIF_F_IP_CSUM); 4928 case htons(ETH_P_IPV6): 4929 return !!(features & NETIF_F_IPV6_CSUM); 4930 default: 4931 return false; 4932 } 4933 } 4934 4935 #ifdef CONFIG_BUG 4936 void netdev_rx_csum_fault(struct net_device *dev, struct sk_buff *skb); 4937 #else 4938 static inline void netdev_rx_csum_fault(struct net_device *dev, 4939 struct sk_buff *skb) 4940 { 4941 } 4942 #endif 4943 /* rx skb timestamps */ 4944 void net_enable_timestamp(void); 4945 void net_disable_timestamp(void); 4946 4947 #ifdef CONFIG_PROC_FS 4948 int __init dev_proc_init(void); 4949 #else 4950 #define dev_proc_init() 0 4951 #endif 4952 4953 static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops, 4954 struct sk_buff *skb, struct net_device *dev, 4955 bool more) 4956 { 4957 __this_cpu_write(softnet_data.xmit.more, more); 4958 return ops->ndo_start_xmit(skb, dev); 4959 } 4960 4961 static inline bool netdev_xmit_more(void) 4962 { 4963 return __this_cpu_read(softnet_data.xmit.more); 4964 } 4965 4966 static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev, 4967 struct netdev_queue *txq, bool more) 4968 { 4969 const struct net_device_ops *ops = dev->netdev_ops; 4970 netdev_tx_t rc; 4971 4972 rc = __netdev_start_xmit(ops, skb, dev, more); 4973 if (rc == NETDEV_TX_OK) 4974 txq_trans_update(txq); 4975 4976 return rc; 4977 } 4978 4979 int netdev_class_create_file_ns(const struct class_attribute *class_attr, 4980 const void *ns); 4981 void netdev_class_remove_file_ns(const struct class_attribute *class_attr, 4982 const void *ns); 4983 4984 extern const struct kobj_ns_type_operations net_ns_type_operations; 4985 4986 const char *netdev_drivername(const struct net_device *dev); 4987 4988 void linkwatch_run_queue(void); 4989 4990 static inline netdev_features_t netdev_intersect_features(netdev_features_t f1, 4991 netdev_features_t f2) 4992 { 4993 if ((f1 ^ f2) & NETIF_F_HW_CSUM) { 4994 if (f1 & NETIF_F_HW_CSUM) 4995 f1 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM); 4996 else 4997 f2 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM); 4998 } 4999 5000 return f1 & f2; 5001 } 5002 5003 static inline netdev_features_t netdev_get_wanted_features( 5004 struct net_device *dev) 5005 { 5006 return (dev->features & ~dev->hw_features) | dev->wanted_features; 5007 } 5008 netdev_features_t netdev_increment_features(netdev_features_t all, 5009 netdev_features_t one, netdev_features_t mask); 5010 5011 /* Allow TSO being used on stacked device : 5012 * Performing the GSO segmentation before last device 5013 * is a performance improvement. 5014 */ 5015 static inline netdev_features_t netdev_add_tso_features(netdev_features_t features, 5016 netdev_features_t mask) 5017 { 5018 return netdev_increment_features(features, NETIF_F_ALL_TSO, mask); 5019 } 5020 5021 int __netdev_update_features(struct net_device *dev); 5022 void netdev_update_features(struct net_device *dev); 5023 void netdev_change_features(struct net_device *dev); 5024 5025 void netif_stacked_transfer_operstate(const struct net_device *rootdev, 5026 struct net_device *dev); 5027 5028 netdev_features_t passthru_features_check(struct sk_buff *skb, 5029 struct net_device *dev, 5030 netdev_features_t features); 5031 netdev_features_t netif_skb_features(struct sk_buff *skb); 5032 5033 static inline bool net_gso_ok(netdev_features_t features, int gso_type) 5034 { 5035 netdev_features_t feature = (netdev_features_t)gso_type << NETIF_F_GSO_SHIFT; 5036 5037 /* check flags correspondence */ 5038 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT)); 5039 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT)); 5040 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT)); 5041 BUILD_BUG_ON(SKB_GSO_TCP_FIXEDID != (NETIF_F_TSO_MANGLEID >> NETIF_F_GSO_SHIFT)); 5042 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT)); 5043 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT)); 5044 BUILD_BUG_ON(SKB_GSO_GRE != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT)); 5045 BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT)); 5046 BUILD_BUG_ON(SKB_GSO_IPXIP4 != (NETIF_F_GSO_IPXIP4 >> NETIF_F_GSO_SHIFT)); 5047 BUILD_BUG_ON(SKB_GSO_IPXIP6 != (NETIF_F_GSO_IPXIP6 >> NETIF_F_GSO_SHIFT)); 5048 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT)); 5049 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT)); 5050 BUILD_BUG_ON(SKB_GSO_PARTIAL != (NETIF_F_GSO_PARTIAL >> NETIF_F_GSO_SHIFT)); 5051 BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT)); 5052 BUILD_BUG_ON(SKB_GSO_SCTP != (NETIF_F_GSO_SCTP >> NETIF_F_GSO_SHIFT)); 5053 BUILD_BUG_ON(SKB_GSO_ESP != (NETIF_F_GSO_ESP >> NETIF_F_GSO_SHIFT)); 5054 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_GSO_UDP >> NETIF_F_GSO_SHIFT)); 5055 BUILD_BUG_ON(SKB_GSO_UDP_L4 != (NETIF_F_GSO_UDP_L4 >> NETIF_F_GSO_SHIFT)); 5056 BUILD_BUG_ON(SKB_GSO_FRAGLIST != (NETIF_F_GSO_FRAGLIST >> NETIF_F_GSO_SHIFT)); 5057 5058 return (features & feature) == feature; 5059 } 5060 5061 static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features) 5062 { 5063 return net_gso_ok(features, skb_shinfo(skb)->gso_type) && 5064 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST)); 5065 } 5066 5067 static inline bool netif_needs_gso(struct sk_buff *skb, 5068 netdev_features_t features) 5069 { 5070 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) || 5071 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) && 5072 (skb->ip_summed != CHECKSUM_UNNECESSARY))); 5073 } 5074 5075 static inline void netif_set_gso_max_size(struct net_device *dev, 5076 unsigned int size) 5077 { 5078 dev->gso_max_size = size; 5079 } 5080 5081 static inline void skb_gso_error_unwind(struct sk_buff *skb, __be16 protocol, 5082 int pulled_hlen, u16 mac_offset, 5083 int mac_len) 5084 { 5085 skb->protocol = protocol; 5086 skb->encapsulation = 1; 5087 skb_push(skb, pulled_hlen); 5088 skb_reset_transport_header(skb); 5089 skb->mac_header = mac_offset; 5090 skb->network_header = skb->mac_header + mac_len; 5091 skb->mac_len = mac_len; 5092 } 5093 5094 static inline bool netif_is_macsec(const struct net_device *dev) 5095 { 5096 return dev->priv_flags & IFF_MACSEC; 5097 } 5098 5099 static inline bool netif_is_macvlan(const struct net_device *dev) 5100 { 5101 return dev->priv_flags & IFF_MACVLAN; 5102 } 5103 5104 static inline bool netif_is_macvlan_port(const struct net_device *dev) 5105 { 5106 return dev->priv_flags & IFF_MACVLAN_PORT; 5107 } 5108 5109 static inline bool netif_is_bond_master(const struct net_device *dev) 5110 { 5111 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING; 5112 } 5113 5114 static inline bool netif_is_bond_slave(const struct net_device *dev) 5115 { 5116 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING; 5117 } 5118 5119 static inline bool netif_supports_nofcs(struct net_device *dev) 5120 { 5121 return dev->priv_flags & IFF_SUPP_NOFCS; 5122 } 5123 5124 static inline bool netif_has_l3_rx_handler(const struct net_device *dev) 5125 { 5126 return dev->priv_flags & IFF_L3MDEV_RX_HANDLER; 5127 } 5128 5129 static inline bool netif_is_l3_master(const struct net_device *dev) 5130 { 5131 return dev->priv_flags & IFF_L3MDEV_MASTER; 5132 } 5133 5134 static inline bool netif_is_l3_slave(const struct net_device *dev) 5135 { 5136 return dev->priv_flags & IFF_L3MDEV_SLAVE; 5137 } 5138 5139 static inline bool netif_is_bridge_master(const struct net_device *dev) 5140 { 5141 return dev->priv_flags & IFF_EBRIDGE; 5142 } 5143 5144 static inline bool netif_is_bridge_port(const struct net_device *dev) 5145 { 5146 return dev->priv_flags & IFF_BRIDGE_PORT; 5147 } 5148 5149 static inline bool netif_is_ovs_master(const struct net_device *dev) 5150 { 5151 return dev->priv_flags & IFF_OPENVSWITCH; 5152 } 5153 5154 static inline bool netif_is_ovs_port(const struct net_device *dev) 5155 { 5156 return dev->priv_flags & IFF_OVS_DATAPATH; 5157 } 5158 5159 static inline bool netif_is_any_bridge_port(const struct net_device *dev) 5160 { 5161 return netif_is_bridge_port(dev) || netif_is_ovs_port(dev); 5162 } 5163 5164 static inline bool netif_is_team_master(const struct net_device *dev) 5165 { 5166 return dev->priv_flags & IFF_TEAM; 5167 } 5168 5169 static inline bool netif_is_team_port(const struct net_device *dev) 5170 { 5171 return dev->priv_flags & IFF_TEAM_PORT; 5172 } 5173 5174 static inline bool netif_is_lag_master(const struct net_device *dev) 5175 { 5176 return netif_is_bond_master(dev) || netif_is_team_master(dev); 5177 } 5178 5179 static inline bool netif_is_lag_port(const struct net_device *dev) 5180 { 5181 return netif_is_bond_slave(dev) || netif_is_team_port(dev); 5182 } 5183 5184 static inline bool netif_is_rxfh_configured(const struct net_device *dev) 5185 { 5186 return dev->priv_flags & IFF_RXFH_CONFIGURED; 5187 } 5188 5189 static inline bool netif_is_failover(const struct net_device *dev) 5190 { 5191 return dev->priv_flags & IFF_FAILOVER; 5192 } 5193 5194 static inline bool netif_is_failover_slave(const struct net_device *dev) 5195 { 5196 return dev->priv_flags & IFF_FAILOVER_SLAVE; 5197 } 5198 5199 /* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */ 5200 static inline void netif_keep_dst(struct net_device *dev) 5201 { 5202 dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM); 5203 } 5204 5205 /* return true if dev can't cope with mtu frames that need vlan tag insertion */ 5206 static inline bool netif_reduces_vlan_mtu(struct net_device *dev) 5207 { 5208 /* TODO: reserve and use an additional IFF bit, if we get more users */ 5209 return dev->priv_flags & IFF_MACSEC; 5210 } 5211 5212 extern struct pernet_operations __net_initdata loopback_net_ops; 5213 5214 /* Logging, debugging and troubleshooting/diagnostic helpers. */ 5215 5216 /* netdev_printk helpers, similar to dev_printk */ 5217 5218 static inline const char *netdev_name(const struct net_device *dev) 5219 { 5220 if (!dev->name[0] || strchr(dev->name, '%')) 5221 return "(unnamed net_device)"; 5222 return dev->name; 5223 } 5224 5225 static inline bool netdev_unregistering(const struct net_device *dev) 5226 { 5227 return dev->reg_state == NETREG_UNREGISTERING; 5228 } 5229 5230 static inline const char *netdev_reg_state(const struct net_device *dev) 5231 { 5232 switch (dev->reg_state) { 5233 case NETREG_UNINITIALIZED: return " (uninitialized)"; 5234 case NETREG_REGISTERED: return ""; 5235 case NETREG_UNREGISTERING: return " (unregistering)"; 5236 case NETREG_UNREGISTERED: return " (unregistered)"; 5237 case NETREG_RELEASED: return " (released)"; 5238 case NETREG_DUMMY: return " (dummy)"; 5239 } 5240 5241 WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, dev->reg_state); 5242 return " (unknown)"; 5243 } 5244 5245 __printf(3, 4) __cold 5246 void netdev_printk(const char *level, const struct net_device *dev, 5247 const char *format, ...); 5248 __printf(2, 3) __cold 5249 void netdev_emerg(const struct net_device *dev, const char *format, ...); 5250 __printf(2, 3) __cold 5251 void netdev_alert(const struct net_device *dev, const char *format, ...); 5252 __printf(2, 3) __cold 5253 void netdev_crit(const struct net_device *dev, const char *format, ...); 5254 __printf(2, 3) __cold 5255 void netdev_err(const struct net_device *dev, const char *format, ...); 5256 __printf(2, 3) __cold 5257 void netdev_warn(const struct net_device *dev, const char *format, ...); 5258 __printf(2, 3) __cold 5259 void netdev_notice(const struct net_device *dev, const char *format, ...); 5260 __printf(2, 3) __cold 5261 void netdev_info(const struct net_device *dev, const char *format, ...); 5262 5263 #define netdev_level_once(level, dev, fmt, ...) \ 5264 do { \ 5265 static bool __print_once __read_mostly; \ 5266 \ 5267 if (!__print_once) { \ 5268 __print_once = true; \ 5269 netdev_printk(level, dev, fmt, ##__VA_ARGS__); \ 5270 } \ 5271 } while (0) 5272 5273 #define netdev_emerg_once(dev, fmt, ...) \ 5274 netdev_level_once(KERN_EMERG, dev, fmt, ##__VA_ARGS__) 5275 #define netdev_alert_once(dev, fmt, ...) \ 5276 netdev_level_once(KERN_ALERT, dev, fmt, ##__VA_ARGS__) 5277 #define netdev_crit_once(dev, fmt, ...) \ 5278 netdev_level_once(KERN_CRIT, dev, fmt, ##__VA_ARGS__) 5279 #define netdev_err_once(dev, fmt, ...) \ 5280 netdev_level_once(KERN_ERR, dev, fmt, ##__VA_ARGS__) 5281 #define netdev_warn_once(dev, fmt, ...) \ 5282 netdev_level_once(KERN_WARNING, dev, fmt, ##__VA_ARGS__) 5283 #define netdev_notice_once(dev, fmt, ...) \ 5284 netdev_level_once(KERN_NOTICE, dev, fmt, ##__VA_ARGS__) 5285 #define netdev_info_once(dev, fmt, ...) \ 5286 netdev_level_once(KERN_INFO, dev, fmt, ##__VA_ARGS__) 5287 5288 #define MODULE_ALIAS_NETDEV(device) \ 5289 MODULE_ALIAS("netdev-" device) 5290 5291 #if defined(CONFIG_DYNAMIC_DEBUG) || \ 5292 (defined(CONFIG_DYNAMIC_DEBUG_CORE) && defined(DYNAMIC_DEBUG_MODULE)) 5293 #define netdev_dbg(__dev, format, args...) \ 5294 do { \ 5295 dynamic_netdev_dbg(__dev, format, ##args); \ 5296 } while (0) 5297 #elif defined(DEBUG) 5298 #define netdev_dbg(__dev, format, args...) \ 5299 netdev_printk(KERN_DEBUG, __dev, format, ##args) 5300 #else 5301 #define netdev_dbg(__dev, format, args...) \ 5302 ({ \ 5303 if (0) \ 5304 netdev_printk(KERN_DEBUG, __dev, format, ##args); \ 5305 }) 5306 #endif 5307 5308 #if defined(VERBOSE_DEBUG) 5309 #define netdev_vdbg netdev_dbg 5310 #else 5311 5312 #define netdev_vdbg(dev, format, args...) \ 5313 ({ \ 5314 if (0) \ 5315 netdev_printk(KERN_DEBUG, dev, format, ##args); \ 5316 0; \ 5317 }) 5318 #endif 5319 5320 /* 5321 * netdev_WARN() acts like dev_printk(), but with the key difference 5322 * of using a WARN/WARN_ON to get the message out, including the 5323 * file/line information and a backtrace. 5324 */ 5325 #define netdev_WARN(dev, format, args...) \ 5326 WARN(1, "netdevice: %s%s: " format, netdev_name(dev), \ 5327 netdev_reg_state(dev), ##args) 5328 5329 #define netdev_WARN_ONCE(dev, format, args...) \ 5330 WARN_ONCE(1, "netdevice: %s%s: " format, netdev_name(dev), \ 5331 netdev_reg_state(dev), ##args) 5332 5333 /* netif printk helpers, similar to netdev_printk */ 5334 5335 #define netif_printk(priv, type, level, dev, fmt, args...) \ 5336 do { \ 5337 if (netif_msg_##type(priv)) \ 5338 netdev_printk(level, (dev), fmt, ##args); \ 5339 } while (0) 5340 5341 #define netif_level(level, priv, type, dev, fmt, args...) \ 5342 do { \ 5343 if (netif_msg_##type(priv)) \ 5344 netdev_##level(dev, fmt, ##args); \ 5345 } while (0) 5346 5347 #define netif_emerg(priv, type, dev, fmt, args...) \ 5348 netif_level(emerg, priv, type, dev, fmt, ##args) 5349 #define netif_alert(priv, type, dev, fmt, args...) \ 5350 netif_level(alert, priv, type, dev, fmt, ##args) 5351 #define netif_crit(priv, type, dev, fmt, args...) \ 5352 netif_level(crit, priv, type, dev, fmt, ##args) 5353 #define netif_err(priv, type, dev, fmt, args...) \ 5354 netif_level(err, priv, type, dev, fmt, ##args) 5355 #define netif_warn(priv, type, dev, fmt, args...) \ 5356 netif_level(warn, priv, type, dev, fmt, ##args) 5357 #define netif_notice(priv, type, dev, fmt, args...) \ 5358 netif_level(notice, priv, type, dev, fmt, ##args) 5359 #define netif_info(priv, type, dev, fmt, args...) \ 5360 netif_level(info, priv, type, dev, fmt, ##args) 5361 5362 #if defined(CONFIG_DYNAMIC_DEBUG) || \ 5363 (defined(CONFIG_DYNAMIC_DEBUG_CORE) && defined(DYNAMIC_DEBUG_MODULE)) 5364 #define netif_dbg(priv, type, netdev, format, args...) \ 5365 do { \ 5366 if (netif_msg_##type(priv)) \ 5367 dynamic_netdev_dbg(netdev, format, ##args); \ 5368 } while (0) 5369 #elif defined(DEBUG) 5370 #define netif_dbg(priv, type, dev, format, args...) \ 5371 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args) 5372 #else 5373 #define netif_dbg(priv, type, dev, format, args...) \ 5374 ({ \ 5375 if (0) \ 5376 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \ 5377 0; \ 5378 }) 5379 #endif 5380 5381 /* if @cond then downgrade to debug, else print at @level */ 5382 #define netif_cond_dbg(priv, type, netdev, cond, level, fmt, args...) \ 5383 do { \ 5384 if (cond) \ 5385 netif_dbg(priv, type, netdev, fmt, ##args); \ 5386 else \ 5387 netif_ ## level(priv, type, netdev, fmt, ##args); \ 5388 } while (0) 5389 5390 #if defined(VERBOSE_DEBUG) 5391 #define netif_vdbg netif_dbg 5392 #else 5393 #define netif_vdbg(priv, type, dev, format, args...) \ 5394 ({ \ 5395 if (0) \ 5396 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \ 5397 0; \ 5398 }) 5399 #endif 5400 5401 /* 5402 * The list of packet types we will receive (as opposed to discard) 5403 * and the routines to invoke. 5404 * 5405 * Why 16. Because with 16 the only overlap we get on a hash of the 5406 * low nibble of the protocol value is RARP/SNAP/X.25. 5407 * 5408 * 0800 IP 5409 * 0001 802.3 5410 * 0002 AX.25 5411 * 0004 802.2 5412 * 8035 RARP 5413 * 0005 SNAP 5414 * 0805 X.25 5415 * 0806 ARP 5416 * 8137 IPX 5417 * 0009 Localtalk 5418 * 86DD IPv6 5419 */ 5420 #define PTYPE_HASH_SIZE (16) 5421 #define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1) 5422 5423 extern struct list_head ptype_all __read_mostly; 5424 extern struct list_head ptype_base[PTYPE_HASH_SIZE] __read_mostly; 5425 5426 extern struct net_device *blackhole_netdev; 5427 5428 #endif /* _LINUX_NETDEVICE_H */ 5429