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